Skip to content

Console reference

This page is the precise reference for the Cora AI console as it is today: what each screen shows, who can do what, and the exact labels and messages. It is checked against the console and the API. A console feature that disagrees with this page is a defect in one or the other (D0017), and the fix goes into whichever is wrong.

The screens follow the left rail (Home, Missions, Heroes, Artifacts, Connections), then Settings, then plans and billing. Decisions are cited by id, for example D0015. Terms are defined in the glossary.

Screens: Sign up (/signup), Verify email (/verify), Sign in (/login), Forgot password (/forgot-password), Set a new password (/new-password), Create your organisation (/create-org), Accept your invitation (/invites/<id>), and the one-time question What should we call you?.

What you see.

These pages are in the Team Cora look, the same as the website (D0900): a cream page, the Team Cora logo at the top (it opens the website) with Website and Docs links, one white card with the page’s form, and on a wide screen a lime panel beside it. Create your organisation and Accept your invitation use the same layout, so a person with no organisation yet sees no console menu. The browser tab names the page and Team Cora, for example “Sign in · Team Cora”. While a button’s request runs, the button is greyed out with a small spinner.

  • Sign up asks for an Email and a Password, under the line “Create your account. Next, you name your organisation.” The hint says “Use at least 10 characters, with an upper-case letter, a lower-case letter and a number.” Sign up sends a code to the address and opens Verify email. “Already have an account? Sign in” links to Sign in. There is no sign-in with a Google account (D0046).
  • Verify email says “We have sent a code to . Enter it here to finish setting up your account. You need to verify your email before you can use the console.” It asks for the Code, with Verify email and Send a new code. Once the address is verified, Sign in opens with “Your email is verified. Sign in to continue.”
  • Sign in says “Welcome back. Your missions are where you left them.” It asks for the email and the password, with Forgot password? and “New to Team Cora? Sign up”. An address not verified yet goes to Verify email. A person invited with a temporary password goes to Set a new password. Anyone else goes to Home, or back to the invitation link that sent them to Sign in.
  • Forgot password asks for the email and Send code, then the Code and a New password, with Change password. It says the same thing whether or not an account exists for that address: “If there is an account for that email, we have sent it a code. Enter the code and your new password.” Afterwards Sign in says “Your password is changed. Sign in with your new password.”
  • Set a new password appears the first time an invited person signs in with the temporary password from their invitation email: “You signed in with the temporary password from your invitation. Choose your own password to continue.” Two passwords that differ say “The two passwords do not match.”
  • Create your organisation is sign-up’s second step. It asks for Your name and the Organisation name, under the line “Name your organisation to finish signing up. You will be its admin and hold its fallback decision role, so every decision and approval can reach someone from the start.” Create organisation opens Home. An empty name says “Give your organisation a name.” Invitations waiting for your verified address come first, each with its organisation, its starting roles, when it expires and Accept; the form then sits under “Or create your own organisation”. A plan chosen on the pricing page before sign-up opens Billing with that plan ticked; nothing is charged until Continue to payment. The card’s footer says “Questions? Read the docs” and has Log out, for someone signed in with the wrong address.
  • Accept your invitation is the page an invitation link opens. You sign in first, and the page comes back. It says “You have been invited to join an organisation on Team Cora. Accepting adds you to it with the roles you were invited with. Only the address the invitation was sent to can accept it.” It asks for Your name, and Accept invitation opens Home. An invitation for another address, or one that has expired, been withdrawn or been used, gets one message, so the page never reveals which (D0048; see State messages). Its footer has the same docs link and Log out.
  • What should we call you? shows once, in place of any page, to someone who joined before Cora asked for names: “The people you work with see your name on missions, in the dialogue and on cards.”

Rules.

  • Someone signed in who belongs to no organisation is sent to Create your organisation, unless invitations are waiting for them. Then Home lists the invitations, with “Not joining? Create your own organisation instead.”
  • An invitation belongs to the email address it was sent to, and only someone signed in with that verified address can accept it. It stays open for 7 days. How invitations are sent, resent and revoked is under Members page.
  • Accepting makes you a member of the organisation with the invitation’s starting roles. Whoever creates an organisation is its first org admin and holds its fallback decision role.

The console has three kinds of rights. Each section below says which one a control needs, and the API checks every change again, so a control the console hides is also refused by the API.

Who What they can do
Org admin: an organisation’s own role. Whoever creates the organisation is its first. Invite and remove people, make someone an org admin or a member, manage the roles, set up the organisation’s connections and switch MCP on or off (Settings, Permissions), change Web access, change Cora’s memory, open Usage and billing, choose and pay for a plan, and switch a hero off or allow it again.
Member: everyone in the organisation. Create missions and heroes, work in the missions they are in, connect their own services and share them in their missions, read the members, the roles and Cora’s memory, and leave the organisation.
Mission admin: whoever created a mission, and anyone they make an admin (D0623). In that mission: add and remove people and make them admins, switch its skills and its web access, manage its vault keys, pause and delete its schedules, delete its metrics, add and remove heroes, and make it a hero. A hero’s admins are its mission’s admins.

Being an org admin does not make you a mission admin. An org admin who is in a mission may also rename it, change its icon colour and its objectives, delete its metrics and manage its vault keys. Everything else in the mission admin’s list needs one of the mission’s own admins: adding and removing its people and making them admins, switching its skills and its web access, pausing and deleting its schedules, adding and removing heroes, and making it a hero (the console still offers an org admin Make it a hero and the API refuses the save, BG-01M7RGV5). Each section below says which.

What you see. There is no top bar (D0717). The console’s navigation is a narrow rail down the left side, 76 pixels wide, with its name under each icon (US-01M7MX5A, D0824; the names used to show only on hover), and each page’s own heading starts at the very top of the window, so the chat gets the whole height of the screen. At the top of the rail are Home, Missions, Heroes (a shield with a star; US-01M7Q2NX, D0856), Artifacts and Connections. There is no People button: the members and the roles are in Settings, under People and roles, opened from the account menu (US-01M7GSCV, D0722). At the bottom are Notices (a bell, with a red count of your unread notices), Help (it opens this docs site) and the account button, a round button with your initials, named with your initials and your name. Home, Missions, Heroes, Artifacts, Connections, Notices and Help each show their name under the icon, and each is at least 44 pixels each way. Screen readers hear the same name. The icon of the page you are on is shaded blue with a bar on its left edge. Missions shows a red count of the missions that have unread messages or need you, and the count is read as part of its name, for example “Missions, 3 need attention”. It is hidden when the count is zero. On a phone the rail becomes a row of the same icons above the page, with no names under them. The browser tab still names the page you are on.

  • Home is where you start: what is waiting on you, how each mission is going and what Cora did recently (see Home below).
  • Missions opens the chat window: the mission list and the open mission’s chat (see Mission chat below).
  • Artifacts opens the organisation’s artifacts at /artifacts (see Artifacts below). It was Files at /files until US-01M7PP0C (D0853); /files still works and brings you to Artifacts.
  • Connections is where each person connects the services Cora may use and shares them in missions, org admins set services up for the organisation, It opens the Connections page in the main area, like Missions and Heroes (see Connections page below; US-01M7SS78, D0882; it was a window over the page from US-01M7GSCV), which links to the guide to connecting an MCP server and, for org admins, to Workspace usage. Web access is in Settings, for org admins.
  • Notices opens the list of your notices (see Notices below); opening one marks it read and goes to its card. A click anywhere outside the list, or Escape, closes it.
  • Help opens the docs site.

The account menu. The account button opens a small menu beside it. Escape or a click outside it closes it, and focus goes back to the button. It holds, in order:

  • Your name, with the organisation and your role in it (“Harbour Lane Digital · org admin”).
  • Billing and plan. On a paid or internal plan it shows the plan’s name and, for an org admin, the percent of the plan’s AI usage used (“Pro · 42% used”). During the free trial it shows Free trial, 9 days left, a meter with 34% of your trial’s AI usage used and Choose a plan, which opens Billing (both for org admins only; members cannot pay). The row opens Billing for an org admin.
  • People and roles, which opens the Settings window on People and roles, at its Members tab. Everyone has it; a member reads the members and the roles there, and an org admin changes them and presses Invite members.
  • Settings, which opens the Settings window (see Settings below).
  • Switch organisation, only when you belong to more than one. It lists your organisations and uses the same choice as the mission list’s switcher.
  • Log out. It first asks “Sign out of Cora?” in a small pop-up beside the account button, with Cancel and Sign out; focus starts on Cancel, and Esc, Cancel or a click outside it keeps you signed in and puts focus back on the account button. On a phone the pop-up stays on the screen.

The dot. A small dot on the account button means the account menu needs a look (D0717). It shows when the free trial has 3 days or less left, when the trial’s usage is at 80% or more (org admins only, because members never see the usage figure), or when the plan is paused. The reason is part of the button’s name for a screen reader, for example “MC, Megan Carvalho, free trial ends in 2 days”. Members see the dot only in the last 3 days of a trial or when it is paused. Nobody sees it on a paid or internal plan that is running.

Rules.

  • The existing addresses keep working: /members, /roles and /connectors. /members was the People page on the rail; it now opens Settings on People and roles, and /members?invite=1 still opens the invite form for an org admin. /projects, /artifacts and /heroes are new. /files goes to /artifacts with the same link: ?doc=<id> opens that document, ?project=<id> opens that mission on its Artifacts tab, and ?path=memory/... opens the memory file in Settings, Cora’s memory (D0853, D0854).
  • The icons. Each is a line drawing in the same stroke. Missions is Cora’s spark on its own, as an outline (US-01M7PG57, D0852); the shield with a star stays the mark of a hero. Artifacts is a page with a folded corner over a second page (US-01M7PP0C). Connections is three small rounded squares and a plus in a two by two grid.
  • /projects/:projectId/files takes you to the mission’s page with mission details open on its Artifacts tab (/projects/<id>?files=open, or ?artifacts=open; D0634, D0718, D0853). A mission’s header has no tabs; its Details button opens and closes the pane, whose Artifacts tab also lists the files shared in the mission’s messages, and whose Memory tab holds Cora’s memory for the mission (D0565, D0566, D0621, D0718, D0830, D0853). The other old addresses keep working too: ?hero=new and ?hero=open open Overview, where Make it a hero and the link to the hero’s page are (D0857), and ?vault=add opens Tools with the Add a key form (D0825).

What you see. Home opens with a greeting and today’s date, such as “Good afternoon, Megan” over “Monday 10 November · 3 things need you today”, with All missions and + Create a mission buttons at the right. Under it are four tiles: Waiting for you (the questions, approvals and unread @mentions that wait on you, in red when anything is overdue, with “1 overdue · 2 due today” under it), Unread messages (“across 3 missions”), Active missions (“1 at risk”) and AI usage this month (the percent of the month’s AI usage used, such as “21%”, with a thin bar; Trial AI usage during the free trial). AI usage is never shown in dollars (BG-01M7QVN5). The usage tile is shown to org admins only; a member sees three tiles. Below are the three cards. Needs you lists each question, approval and @mention waiting on you across your missions, overdue first, then by due date, then oldest first. Each has its mission’s avatar and name, a kind tag (Question, Approval or Mention), its title, when it is due and one button: Answer for a question, Review for an approval and Open for a mention. The button opens that mission with the message focused. Missions at a glance shows each active mission you are in with its stage, its health as Cora last reported it (On track, At risk or Off track, or “No status yet”), the next milestone with its date and the number of unread messages. A click opens the mission. Latest from Cora lists her recent updates and documents across your missions, newest first, each linking to the message. Each shows the start of her message as plain text, without Markdown marks such as ** or #, cut to two lines. Home uses the whole width of the window, with the same side margins as the other pages (US-01M7SS78, D0881; it was centred and up to 1,240 pixels wide until then). On a wide screen (1,440 pixels and up) the three cards sit side by side: Needs you, Missions at a glance, Latest from Cora. On a laptop there are two columns, with Latest from Cora under Needs you and Missions at a glance on the right. On a phone the columns stack into one, and the four tiles stretch with the page. Long lines of text in Needs you and Latest from Cora stay at a comfortable reading width inside their cards.

What you can do. Press Answer, Review or Open to go straight to the item. Press a mission to open it. Press All missions for the full list, or + Create a mission. Home reads again when you return to the tab and whenever something changes for you, so the numbers stay current without a reload.

Rules. Everything on Home is yours: only missions you are a current member of, and only the questions, approvals and @mentions that are for you (US-01M71QQ3). A mention stays in the list until you read that mission’s messages. The numbers use the same tests as the mission list’s “needs you” count, so they agree. Spend comes from the organisation’s cap for this month and is for org admins only; the answer for a member has no spend at all. The status Cora reports is a structured record (health, stage, next milestone and a one-line summary) that she sets with her daily update and whenever any of them changes; Home shows the latest. A person who is in no organisation is sent to create one, and invitations waiting for them still show first.

States. Loading, nothing needs you (one short line with a check mark: “You’re all caught up. New questions, approvals and mentions show here.”), no active missions, no updates from Cora yet, and an error with a message.

What you see. Notices at the bottom of the rail opens your list of notices beside the rail (US-01M6AZYS; the list since US-01M7SS78, D0884). It is never taller than the window: the list scrolls inside it, and its header stays put. On a phone it opens under the row of icons, inside the screen. At the top are the heading Notices, Mark all as read (while anything is unread) and two buttons, All and Unread (with how many are unread). The notices are grouped by day, newest first: Today, Yesterday, then the date, such as “16 Nov”. Each row shows a small icon for its kind, what happened (“A question for you”, “An approval waiting for you”, “You were mentioned”, “You were added to a mission”, “A connection is failing” and the rest), the mission it is in with the mission’s spark tile, and a short time: “2 h ago” today, the clock time on earlier days. The full date and time show when you hover over the time. The mission’s name comes from your own missions list; a notice from another organisation shows no mission. Unread notices have a dot and bold words, and a screen reader hears “unread”; read ones are quieter. Notices of the same kind in the same mission on the same day show as one row with how many there are and the time span, for example “You were mentioned · 12” and “11:29 pm to 11:52 pm”. At the foot, “Showing the latest 50” and Show older, which reads the next 50; it goes when there are no more.

What you can do. Press a row to go where it points: its card in the mission, or the page it is about. That marks it read; a row of repeats goes to the newest and marks them all read. Mark all as read marks every unread notice of yours read at once. Unread shows only unread notices, All every one. A click outside the list, or Escape, closes it.

With the keyboard (D0886). Focus is never left on a control that has gone. After Mark all as read, focus is on the list’s heading. After Show older, focus is on the first notice it added, so you carry on reading from there (on the heading when it added none that show). A screen reader hears what happened: “All notices marked as read.”, “2 older notices added.” or “No older notices.” Escape closes the list. If focus was in the list, on the Notices button or nowhere, it goes back to the button; if you had moved on to something else, such as the message box, it stays there. When Escape first closes something of its own, such as the message box’s @ list, the notices list stays open until the next Escape.

Rules. You see only your own notices, in organisations you are still a member of, and Mark all as read changes only yours (POST /v1/me/notifications/read-all). A notice names what happened and where, never what anyone wrote (D0114, D0445).

Screens: Missions, with the mission list on the left and the open mission’s dialogue beside it. On a phone the list is hidden and the home page lists your missions. The chat window is described in the mission dialogue.

What you see.

  • The mission list. The organisation’s missions you are in, most recent message first. Each shows the mission’s spark tile, a rounded square 40 pixels wide in the mission’s colour with Cora’s white spark (US-01M7PG57, D0849; no mission shows a mini hero, D0850; see the hero world), and its own name, who sent the last message and its first line, and its time, which never wraps onto a second line (“8:03 pm”, “18 Nov”). A mission with messages you have not read shows one unread count on the right (the messages from other people since you last read it; “99+” past 99), read aloud as “3 unread”, and its name is bold. A mission with a question or an approval waiting on you says Needs you, read aloud with the number (“Needs you: 2 waiting on you”). There is no dot and no count on the mission’s picture any more (US-01M7MX5A, D0823). Missions are grouped under Missions (the active ones, a hero’s mission with its Hero chip among them), then Finished; a group with nothing in it is not shown. Heroes have their own page on the rail, so the list has no Heroes group or filter (US-01M7Q2NX, D0856; see Hero). A hero’s row shows a small chip beside its name, the hero’s pixel face and “Hero: ” (for example “Hero: Kit”), its time like any other row, and under its name how many missions it helped this week, for example “Helped 12 missions this week”: the different missions that called it and got an answer in the last seven days (“No missions yet this week” for none; US-01M7NQ0S, D0837). Search still finds a hero by its @handle. At the top of the column is the heading Missions with a New mission button beside it (the words only); under them the organisation switcher (when you have more than one), search by name with the Fold mission list icon button beside it (it becomes Expand mission list when folded), and the filters as a row of round buttons: All, Unread and Needs me. Each filter but All says how many missions it would show, for example “Unread 2”; a filter with none shows no number, and the numbers count the whole list, not the search (US-01M7NQ0S, D0836). On a mission page the list sits 16 pixels from the rail (other pages keep their 40-pixel margin), is 296 pixels wide until you drag its edge, and keeps 12 pixels clear before that edge (US-01M7SS78, D0881). A width you chose by dragging is kept. A stored width of exactly 280 pixels, the old default, counts as not chosen and becomes 296: the console used to store it whenever you resized mission details. From now on each edge you move stores its own width only (D0885). Folded, the list is a 64-pixel strip 8 pixels from the rail, with each mission’s spark tile at 32 pixels centred under a 32-pixel Expand mission list button; unread missions keep their dot and the mission you are in its ring. The mission you are in has a clear blue background, border and bar on its left edge. Every icon button shows its name in a dark label on hover and on focus, and screen readers hear the same name (D0552).
  • The header. The mission’s spark tile (48 pixels) and its own name on the left, with the chip “Hero: ” beside the name when the mission is a hero (see Hero). There is no stage chip and no mini hero (US-01M7PG57, D0850). The mission’s admins (a mission admin or an org admin) see Rename beside the name, named “Rename the mission”: it turns the name into a box with Save and Cancel (Esc cancels); a name is 1 to 200 characters, an empty one says “Give the mission a name.”, and the missions list follows at once (D0852). Under the name is one line: “Coordinated by Cora · 3 people · objectives: Launch the autumn range on 1 March.” (Cora is the mission’s coordinator; the count is the people in the mission; a long line is cut short and the objectives show in full on hover; D0850, D0851). The names are in the People pop-up. The page reads the mission again when Cora finishes a run (US-01M7MX0B). At the top right are the People button (the word People and a small chevron; named “People, Cora and N people”; it opens the People pop-up described below) and the Details button (the word Details beside a panel icon, named “Mission details”), which opens and closes the mission details pane and shows that it is open with aria-pressed (D0715, D0718, D0838). The header has no tabs: what a mission holds is in the pane. The header stays at the top and the dialogue fills the rest of the screen height, with no title above it: the messages scroll inside and the message box stays at the bottom (D0565, D0566, D0617).
  • Mission details (the pane). A pane docked on the right of the chat, headed Mission details, with a close button and five tabs on one line, in this order: Overview, Artifacts, Memory, Skills and Tools (US-01M7Q2DK, D0857, which took the Hero tab out: a hero has its own page; US-01M7PP0C, D0853; first US-01M7MXD1, D0825; the pane itself US-01M7FGDN, D0718). They took the place of seven tabs, which wrapped onto two lines. Each tab is 44 pixels tall, and the five share one line at every width the pane can have. It is not a pop-up: the chat, the message box and the mission list stay usable while it is open. The tabs work from the keyboard: the arrow keys move between them, Home and End jump to the first and last, and only the chosen tab is a Tab stop. A tab reads what it shows only the first time you choose it. The pane is closed until you open it with the Details button, so the chat has the full width until you ask for the pane (BG-01M7GZVF, D0725); the first time, it opens on Artifacts. From then on the console remembers, per person in this browser, whether you left it open, its last tab and its width; nothing from the chat is kept. A tab remembered under one of the old names opens the tab that holds it now: About opens Overview, Connections and Vault open Tools, and Files opens Artifacts; Memory opens Memory, and Hero (the tab that was there before US-01M7Q2DK) opens Overview. If you used the console before this change, the pane starts closed once whatever it was before (until then it opened by itself, so an open pane was not a choice); open it and it stays open from then on (D0727). A link that asks for the pane, such as a document card’s Open, still opens it. It starts 560 pixels wide, or as wide as the window allows (US-01M7GSCV, D0723); a width you chose yourself is kept, 320 pixels included. While it is open on a desktop the mission list folds to a slim strip of avatars that you can expand again. On a desktop you can drag the edge between the mission list and the chat, and the edge between the chat and the pane, to change their widths (the list from 200 to 420 pixels, the pane from 240 to 560, and the chat never below 360). The edges can be focused and moved 16 pixels at a time with the arrow keys; double-click, Enter or Home puts a width back to its default. On a phone, or any window 900 pixels wide or less, the pane opens as a full-screen sheet with the same five tabs on one line; it starts closed. While it is open the page behind it cannot be reached: Tab and Shift+Tab stay inside the sheet, and Escape closes it. Closing it puts focus back on the Details button.
    • Overview (US-01M7MXD1, D0826). At the top, Mission name: the mission’s spark tile and its name, with Edit (named “Edit the mission name”) and Change the icon colour for the mission’s admins (US-01M7PG57, D0849, D0852). Change the icon colour shows the tile in all 12 colours as a group named “Icon colour”, each read by its colour’s name, with Save (enabled once the colour differs) and Cancel. Then the mission’s Objectives, with Edit for the mission’s admins (a mission admin or an org admin; a member sees no Edit), named “Edit the objectives” for a screen reader. Edit opens a box with Save and Cancel; the objectives are 1 to 4000 characters, and saving empty ones says “Write the objectives in a few words.”. A mission with none says “No objectives yet.” (D0851). Then Becoming a hero: one line (for example “This mission is getting going. Two more steps and it is ready to be a hero.”) and the four steps before it can be a hero, a list named “Steps before it can be a hero”, each marked done or to do in words a screen reader hears: “Objectives written on ”, “Cora finished her first piece of work”, “Its own skill or connection” (with how many so far, such as “1 skill so far”) and “Ten finished pieces of work” (“3 of 10”). A mission that is ready says “This mission is ready to be a hero.”; its admins see the button Make it a hero, which opens the Make it a hero dialog (hero name, what it is good at, when Cora should call it in, and what stays the same), and everyone else “A mission admin can make it a hero.” An org admin in the mission who is not one of its admins also sees the button today, and the save is refused (BG-01M7RGV5). A hero says “This mission is a hero. Other missions can call on .” with the Hero chip and Open , a link to the hero’s own page (US-01M7Q2DK, D0857; see the hero world). There is no mini hero and no stage name in Overview (D0850). See the hero world. What Cora can use here is a plain list of five lines, each saying what is on and ending in one control that opens the place where it is set: Web search and browser (On or Off, Change), Skills (how many are on, Manage), Connections (how many are on, or “None yet”; Manage or Add), Heroes helping (their names, or “None yet”; Manage or Add, named “Manage heroes” and “Add a hero”) and Vault keys (how many, or “None yet”; Manage or Add). Skills opens the Skills tab; the others open the Tools tab on the section the line is about, which takes focus. A line that cannot be read says “Not known” and still opens its tab. The list took the place of the three At a glance cards. Schedules lists the work Cora does on a timer in this mission (US-01M7JA10): what she does, how often, when it runs next (or that it is paused or done, and why it paused) and who asked. Only the mission’s own admins see Pause or Resume and Delete beside each schedule; an org admin who is not one of the mission’s admins sees the list like any member. Delete asks first, and if it fails the row goes back to its buttons with the reason. A line under the list says what the last change did. A member sees the list with no buttons, and can ask Cora in the mission to pause any schedule. With no schedule it says how to ask Cora for one. See Schedules. Metrics follows the schedules (D0772): for each metric Cora records in this mission, its name, its latest value with its unit, how it changed from the value before in words (for example “up 0.3 points, better”), what it means and where it comes from, a small trend chart of the last 30 values with a text description for screen readers, and a Values list of each date and value. The mission’s admins, and an org admin in the mission, see Delete on a metric and on each value, each asking first; a member sees no buttons. With no metric it says how to ask Cora for one. See Mission metrics. Usage shows one number, the mission’s total as “ tokens so far”, and when and by whom the mission was created. There is no list of people: the People button covers people, and the Tools tab the heroes.
    • Artifacts (US-01M7PP0C, D0853; it was Files, US-01M7MXD1, D0830): one page with a search box, Search this mission’s artifacts, and two parts one under the other, each with its count and one line saying what it is. Documents: “What Cora wrote or saved for the team.” Shared in messages: “Files people attached in the chat.”, with who shared each, when, its size and a Download button. The search narrows both by name as you type, opening the folders that hold a match; a part with no match says so. A picture Cora saved shows a small thumbnail and the word Picture on its row (at most 12 thumbnails are read; the rest keep a plain box). Documents keeps its Refresh button and follows new files as they arrive. A file you choose opens in the middle in the chat’s place, with a close button and Esc, and the message box stays.
    • Memory (US-01M7PP0C, D0853): Cora’s memory for this mission, with CORA.md first, marked Always read, under the line “Notes Cora keeps about this mission. Only she writes here; ask her in the chat to change something.” There is no Upload or New folder: only Cora writes it. It keeps its Refresh button, follows new notes as they arrive, and reads the memory only the first time you choose the tab. A file opens in the chat’s place, as in Artifacts.
    • Skills: the mission’s skills in groups by tag, with the number that are on in the tab’s name (see “Mission skills” below).
    • Tools (US-01M7MXD1, D0828): what Cora can reach from this mission, in four sections: Web, Connections, Heroes helping here and Vault keys. It holds what the Connections and Vault tabs held, and the heroes a mission added. Changes save as you make them.
      • Web: the switch “Let Cora use the web in this mission” with its two lines (see “Web access” below).
      • Connections: one card per kind of connection, then The organisation’s connections with a switch each, and Manage all, which opens the Connections page (see “Connections in mission details” below). It is written in plain words: the pane never says MCP. When connections are switched off for the organisation it says “No connections yet” and “Connections let Cora read and draft in your other tools, such as Analytics or your website. They are switched off for this organisation.”, with Turn them on for an org admin (it opens the Connections page) or Ask an org admin to turn them on for a member.
      • Heroes helping here: the heroes added to this mission (US-01M7F9WE, D0719), each drawn as its mini hero with its hero’s name, its card’s name and what it does, and “Switched off” or “Switched off by an org admin” when it cannot be called now. Mission admins see Remove beside each, which asks first (“Remove Kit from this mission? Cora can no longer call it here until a mission admin adds it again.”), and Add a hero, which opens the organisation’s heroes that are on, without the mission’s own and those already added. Everyone else sees the list only. With none it says “No heroes are helping yet.”, and if they cannot be read, “The heroes could not be loaded.” A line under it says what a hero is, and that Cora can call one in when its skills fit the job. Add a hero is the list’s last line. This is the one place heroes are added and removed: the People pop-up lists people only (US-01M7NQ0S, D0844).
      • Vault keys (US-01M7JEQK, D0738, D0739): a list of the mission’s keys by name, each with its access (Ask first, Allowed or Off), its note, and one line such as “Sent only to www.googleapis.com · added 18 Nov 2026 by Megan · never shown again”; nobody can see a value. Mission admins, and org admins in the mission, see Add a key beside the heading, and beside each key Replace, an access menu, Edit for the note and hosts, and Remove, which asks first. The add form (name, value, allowed hosts, note and access, Ask first by default) is not on the page until Add a key is pressed; focus moves into it, and Cancel or Escape closes it and puts focus back on Add a key. The value field is a password field: it is never filled in and is cleared after you save, whether the save worked or not. A member sees the list with no controls and a line saying a mission admin adds keys. An empty list says what it is for. See The mission vault.
  • Mission skills (the Skills tab). The Skills tab of mission details is named with the count, for example “Skills 4”, heard as “Skills 4 of 63 on” (US-01M7MXD1, D0827). Its skills are in groups by tag, in this order: This mission (skills installed for this mission only), Built in (Cora’s own: her weekly status report, and the weekly scenario skills such as the PageSpeed Insights report, which needs a Google key in the mission’s vault), Organisation (skills installed for every mission), then one group for each library by its plain name: Marketing skills (the 53 skills of the marketingskills library) and Anthropic (the 8 marketing skills of Anthropic’s knowledge-work plugins). A group with no skill is not shown. Each group is a button that opens and closes, with its count on the right, for example “4 of 58 on”. This mission and Built in start open; the library groups start closed, and a closed group still names the skills that are on (“On: Copywriting, Email sequence”); a group with none on shows only its name and count. At the top are a search box, which says how many skills it searches, and two round buttons, All and On, whose On carries the number that are on. Both work across the groups: a group with a match opens and shows only its matches, a group with none is hidden, “No skill matches.” shows when nothing does, and clearing the search puts the groups back as you left them. Each row has its switch at the front, then the title in bold, then a two-line description (the full text on hover); the whole row is clickable for admins. A skill waiting for approval (one Cora proposed, see skills Cora proposes) is listed above the groups: “ is waiting for approval”, where it came from and how many files, and Review, which opens the same card the chat shows, in place: read every file, then Approve and install or Decline for a person who may answer it. Under the groups a line says how to add a skill: ask Cora in the chat with a GitHub link, or ask her to write one. Only the weekly status report is on by default, including in existing missions; the library skills are off until a mission admin switches them on (a new mission shows “1 of 62 on”), and a skill added later is off unless it is Cora’s own (cost, D0631). Mission admins can switch skills (optimistic; if the save fails the switch goes back to the last state the server confirmed, with a message); members see disabled switches and the note “Only mission admins can switch skills on or off.” A skill that is off is not offered to Cora’s run and cannot be read by her; Cora is told how many skills are off (not their names); asking for one by name that she cannot see gets an answer that it may be switched off and that a mission admin can turn it on. There is no skill picker in the message box: people ask Cora by name in plain words. While Cora works, the thinking line shows “Using ”; for each of her replies written in a run that used skills, How I worked on the reply lists a chip per skill, with the description on hover and as its accessible description (US-01M6XQQF, D0630, D0631).
  • Web access in the dialogue and mission details (US-01M71PTP, D0637). While Cora searches the web, the “Cora is working” line adds “Searching the web: ” or “Reading ” (for about half a minute after each web call). For a Cora reply written in a run that used the web, How I worked on the reply (US-01M7MX5A, D0821) opens a chip such as “2 searches · 4 pages” and a short list headed “Pages Cora looked at”: each page’s site and title as a link that opens in a new tab (rel="noopener noreferrer", only for http and https pages), with when it was read, or “Could not load” for a page that failed, and a note that web pages are third-party content read as information only. The titles are shown as text. The Tools tab of mission details has a Web section (US-01M7MXD1; it was in the Connections tab): a switch “Let Cora use the web in this mission” that a mission admin can change (members see it disabled with “Only a mission admin can change this.”), and its note “Searches count towards the organisation’s monthly limit.”, with two lines under it that follow it: “Search and read pages” and “Use a browser, for screenshots and clicking through pages” (unticked, with a note, when an org admin has turned the browser off). While the mission’s switch is off both are greyed and unticked and say “Turn on web access first”. When the organisation has web access off, there is only a line saying so and that an org admin can turn it on in Settings › Web access. Nothing here is kept in the browser.
  • Connections in the Tools tab (US-01M7FGAJ; mockup screen 3, without its tool lists, D0716; moved from the Connections tab by US-01M7MXD1). One card per kind of service, as a plain list with no bullets (a screen reader still hears a list): its name, and the access set on the connection that is on here, for example “19 tools: 16 allowed, 3 ask first”. When one is shared here the card says whose and when (“Your connection · shared 12 Nov 2026” or “Megan’s connection · shared 12 Nov 2026”), with Stop sharing for its owner and for an org admin. A connection an org admin made available to the whole organisation reads “Available to the whole organisation (Megan’s connection). Switch it on for this mission.” with an On/Off switch. When none is shared the card says “Not shared here yet” (“. You have no connection” is added when you have none), with Share mine for a person who owns one, or Add one, which opens the Connections page. Manage all, beside the Connections heading, opens the Connections page. There is no tool tick-list anywhere. Below the cards the sections described next hold the same controls in full.
  • Connections in mission details (US-01M746ZB, D0643, D0644). The Tools tab of mission details has a Connections section: “Connections Cora can use here. One of each kind per mission. Only the owner can share or stop sharing.” One row for each kind (for example Google Analytics 4): when one is shared, “’s connection · shared ” and Stop sharing (shown to the owner, and to an org admin); when none is, “Not shared here yet” and Share mine for a person who owns one, or Add one (which opens the Connections page) for a person who does not; a person whose connection cannot be shared because a colleague’s is already shared sees “’s can’t be shared here: this mission already uses ’s. Only can stop sharing it.” A connection an org admin made available to the whole organisation shows “Available to the whole organisation (’s connection)” with an On/Off switch that any mission member can use (a mission that already uses a connection of that kind, shared by a person, shows whose it is and cannot switch the other on). A kind the operator has not set up is not listed. Any mission member can read the section. Under it, The organisation’s connections lists the organisation’s connectors with one On/Off switch each (for an org admin or an admin-role holder) and the connection’s access, such as “19 tools: 16 allowed, 3 ask first”; there are no tool tick-lists. A mission only switches a connection on or off, or shares and stops sharing a personal one: what Cora may do with each tool is set once on the connection, on the Connections page (D0716).
  • The People button (at the top right of the header; US-01M73M62, moved there by US-01M7FGDN; a button with the word People since US-01M7MX5A). It opens, under the button (on a phone it spans the screen width), a popover headed “People (N)”. Cora is first and cannot be removed. Then come the members, with an Admin badge on mission admins and “(you)” beside your own name. Every member sees Leave, which asks “Leave this mission?” in the popover before it does anything; the last mission admin cannot leave, so the button is greyed out with the hint “You’re the only admin. Make someone else an admin before you leave.” Mission admins also see an × beside each person (not Cora, not themselves) when they hover over or tab to a row, which asks before it removes the person; a small Make admin button beside each person who is not an admin; and Add people at the bottom. Add people opens a dialog that searches the organisation’s current members, lets an admin pick several as chips and adds them with Add. Adding someone who is not yet in the organisation by email is not part of this (D0623). Each person added gets a notice in the bell (“You were added to a mission”) and an email that links to the line in the chat; a person made an admin gets “You are now a mission admin”. Taking admin away sends nothing (BG-01M7K3P8, D0780).
    • Heroes are not listed here (US-01M7NQ0S, D0844). The heroes added to the mission are listed, added and removed in the Tools tab of mission details, under Heroes helping here (see above). A hero must be added before anyone can @mention it or Cora can call it in here. Adding or removing a hero writes a quiet line in the chat, such as “John added the hero @seo-audit to the mission.” If you @mention a hero that is not added, a line under the message box says a mission admin can add it in mission details, Tools.
  • Lines in the chat. Adding, removing and leaving each leave a quiet centred line with a person icon in the chat: “John added Sergio to the mission and shared all mission history.”, “John removed Henry from the mission.” or “Henry left the mission.” These lines hold names only, never anything from a private chat. Two changes of the mission’s stage leave a small line too, with the mission’s spark tile: becoming ready to be a hero (“Autumn launch is ready to be a hero. It has ten finished pieces of work and a skill or connection of its own.”) and becoming one (“Autumn launch became a hero. Other missions can now call on Kit.”). It is worked out from the mission’s own work each time and is not a stored message (D0819, D0850; see the mission dialogue).
  • Key requests. When Cora needs a key she does not have, she posts a card with the name she wants, the host it is for and why (request_secret, D0744). A mission admin, or an org admin in the mission, sees Add it in the Vault, which opens mission details on the Tools tab with the Add a key form open and filled with the name and host; everyone else sees that a mission admin can add it. The card never holds a value.
  • The chat. Newest message at the bottom, your own messages on the right, everyone else’s on the left, Cora’s question and approval cards inside its bubble (a message carries no tag naming its kind, such as Question or Update; a question or approval card says what it is in its own words, and screen readers hear “Question from Cora” or “Approval request from Cora”; BG-01M7K3HV), a one-line quote on each reply, Show earlier messages at the top and New messages when something arrives while you are reading above (D0544).
  • The messages. Your own messages sit on the right and your teammates’ on the left, in a soft tint with dark text. Cora’s replies are plain text beside her avatar, with no box; cards inside a reply (documents, questions, approvals, emails, key requests and hero cards) keep their own border. Each message shows its time, and a line with the date separates the days. The messages and the message box sit in one centred column at most 760 pixels wide, whether mission details are open or closed (US-01M7MX5A, D0824).
  • The message box. The text wraps at the full width of the box; when it is empty it says “Message Cora or the team. Type @ to mention someone or call a hero.” Below it, in one row inside the box (US-01M7NQ0S), are three matching soft round-ended buttons, 34 pixels tall with no outline (US-01M7SS78, D0881): a paperclip for Attach files, Thinking, which shows only the choice (“Balanced ▼”) and is read as “Thinking: Balanced”, and the Polish button (the word Polish beside a sparkles icon; “Let Cora tidy your wording” on hover). Then the hint “Shift + Enter for a new line” (“Ctrl + Enter to send · Enter for a new line”, or Cmd on a Mac, when Send with Enter is off) and the round Send button, grey while the box is empty and in the chat’s accent colour once there is something to send. On a phone the hint is hidden and the tools stay on one line. After a polish the box holds Cora’s wording and shows an Undo polish button at the left of the bottom row that puts back what you had typed. There is no token count in or under the box. Enter sends; Shift+Enter starts a new line. Typing @ opens a list of suggestions just above the box (see “Tag people with @” below).

What you can do.

  • Send a message as you typed it, in your own name; it is published at once and never edited (D0545, D0564).
  • Tag people with @. Type @ in the message box and a list opens with the mission’s current people (avatar, name and email), Cora, and Everyone (“Notify everyone in this mission”). Keep typing to narrow it. Arrow Up and Arrow Down move through it, Enter or Tab picks the highlighted line, Esc closes it, and a click picks a line. The pick puts “@Name “ in your message. Anyone you tag gets a notice in the bell and an email that only links to the message; tagging Everyone notifies each person in the mission except you. Tagging Cora works as it always has: Cora replies. In the chat a tagged name is highlighted, and a tag of you stands out more. Only people who are in the mission can be tagged; the list offers no one else, and the server refuses a tag of anyone who is not (D0632).
  • Polish a message with Cora: its wording replaces your text in the box, you can edit it, Undo polish, or send it (D0617). See polishing a message with Cora.
  • Answer a question card with one tap, or press Answer in chat to answer in your own words through the same message box.
  • See a document card when Cora has written a document: a document icon, its name, its place (“Documents · Reports”, or “Documents · Organisation” for an organisation document) and Open in the same row (US-01M7NQ0S, D0839). Press Open to go to that document: a mission’s in the mission’s Artifacts tab, an organisation’s on the Artifacts page (/artifacts?doc=<id>, D0853).
  • See what a hero did: when Cora calls in a hero, its card shows the hero’s mini hero, “ · ” and one line with the result and what it was asked (“Done in 2 minutes · SEO check of the launch page”, or “Could not answer”, “Working”). Press What Kit did (the hero’s name) to open the request, the answer, why it stopped and the time taken (no cost: AI usage is never shown in dollars, BG-01M7QVN5); press it again to close it (D0839).
  • Pause or delete a schedule from the chat: when Cora sets up a schedule, her message carries its card with what she will do, how often, when next and who asked (“Every Monday at 9:00 am · next 23 November · asked by Megan”). The mission’s admins see Pause (or Resume) and Delete, which asks first; everyone else sees the card only. A schedule deleted since says “This schedule was deleted.” (D0841).
  • Act on a next step: when Cora says something is switched off or missing, her reply can carry a button to the place to fix it, shown only to the people allowed to do it (US-01M7NQ0S, D0840; everyone else reads the reply as text). Turn on web access opens the Tools tab at Web (mission admins) or, when web access is off for the organisation, Settings › Web access (org admins); Open Skills opens the Skills tab (mission admins); Add a key opens the Tools tab with the Add a key form (mission admins); Turn on connections opens the Connections page (org admins).
  • Ask an org admin to turn connections on: when connections are switched off for the organisation, the Tools tab’s Connections section has Ask an org admin to turn them on. Each org admin gets one notice in the console, “Someone asked to turn on connections”, which opens the Connections page; pressing it again adds no second notice until they have read the first. The button then says “Your org admins were asked. They see it in Notices.” An org admin sees Turn them on instead (D0842). An organisation document’s card says everyone in the organisation can read it (D0571).
  • See an email card when Cora has emailed people from the mission (D0771): the subject, the people it went to, the names of the files attached and who asked for it. If someone was skipped because an earlier email to them bounced or was marked as spam, the card names them. Only members of the organisation can get Cora’s emails. See Email from Cora.
  • Vote on a choice card when Cora asks the mission to choose (D0773). It shows the question, each option with how many have picked it, an Open link for an option that has a document (a mockup, say) and when it closes. Press Vote on an option to pick it; your pick is marked, and you can pick another until it closes. A mission admin, or an org admin in the mission, also sees Close, which asks first. A closed card shows the result, the winner or a tie, and no buttons. See Choice cards.
  • Add up to 5 files a message, each up to 25 MB, of the seven allowed types. They are kept with the mission, seen only by its members, served only as downloads and not scanned for malware (D0548, D0562). See upload types and limits.
  • Choose how hard Cora thinks about her reply to your message with Thinking. It opens a small menu that says what each choice does: Quick, “Fast answers for simple asks” (the Haiku model), Balanced, “The default for most work” (Cora picks the model for the job; it was called Auto) and Thorough, “Slower, for hard thinking” (the Opus model). The arrow keys move in the menu, Enter chooses, and Escape closes it with focus back on Thinking. Anyone in the mission may choose Thorough, and the box says it is the slowest and uses the most AI usage. The organisation’s AI usage for the month still applies: the reply counts against it like any other. It goes back to Balanced when you reload (D0563, D0824).
  • See how long Cora is working: a running timer on the “Cora is working” line. See how a reply was made: press How I worked on a reply of Cora’s to open “Thought for 1m 20s · 64.5k tokens”, the skills she used, her searches and pages, and the pages as links; press it again to close it. A reply with no recorded run has no button. See what the mission has cost: “1.24M tokens so far” is in mission details’ Overview, under Usage, the one place the total is shown, and it moves whenever the chat reads the mission’s usage. Numbers under 1,000 show as they are, then 12.3k, then 1.24M. Every mission member sees them (D0622; US-01M7MX5A, D0821).
  • Pause, resume or delete a schedule in mission details: the line that says what happened (“The schedule is deleted.”) is announced politely and clears itself after five seconds, so it never stays beside “No schedules yet”. A refusal stays until your next action (US-01M7MX5A).
  • Cora does not tag the person she is answering. She tags a person only when someone other than the asker needs to act or know, or when she comes back to someone more than an hour after their message (D0822).
  • Watch the chat update on its own. A new message, a card change or Cora starting work appears within about two seconds, with no reload. The page is told only that something changed, never what it says, and then reads the change the way it always does, so you see only what you are allowed to see. The mission list’s counts and the notices badge update the same way. If the live connection is down, the page reads again every minute, when you come back to the window and as soon as the connection returns, so nothing is lost (D0557, D0558).

States. No missions yet. No messages yet. Cora is working (a spinning icon and a timer while she works). How I worked open or closed on a reply. Cora paused by the spend cap. Polishing, and a polish you can undo. The @ list open, filtered, and with no match (it closes). A file uploading, ready, or refused (wrong type or over 25 MB).

Screen: New mission (/projects/new), opened from New mission in the mission list or + Create a mission on Home.

What you see. The heading New mission, with “Cora, your mission coordinator, starts as soon as you create it.” Then one card, What is the mission for? (“Cora writes up your objectives in full and makes a first to-do list.”), with:

  • Organisation, a choice shown only when you belong to more than one;
  • Mission name, up to 200 characters: “You can rename it later in mission details.”;
  • Mission objectives, up to 4000 characters: “What should the mission achieve, and by when?”;
  • People, a tick box for each other person in the organisation, or “Nobody else is in the organisation yet. You can add people later.”, and under them “You are the mission admin. You can make others admins later.”

Create mission creates it and Cora starts work. Cancel goes back to Home. There is no hero switch here (D0852): see Hero for making a hero.

Rules. Any member can create a mission, and whoever creates it is its admin. On the free trial a 4th active mission is refused (see State messages).

Screens: the Heroes page on the left rail (/heroes), each hero’s own page (/heroes/<id>) and the Create a hero form (/heroes/new) (US-01M7Q2NX, US-01M7Q2DK; D0856, D0857; designs sdlc-studio/designs/missions-and-heroes/Heroes.dc.html and NewHero.dc.html); Make it a hero in a mission’s Overview; and, for org admins, the Heroes block of Settings, Permissions (US-01M7DRX7, BG-01M7HDK9; RFC-01M7C9F6, D0691, D0693, D0694, D0704, D0707, D0728). The Connections page shows connections only: it has no Heroes group and no hero page any more.

What a hero is. A hero is a specialist any mission can call in, by its own name: you call in Kit with @kit. It is created on the Heroes page, with a training room behind it, or made from a mission that has proved itself. Either way a mission is behind it, and that mission’s admins are the hero’s admins (see the hero world). On the way to becoming a hero a mission goes through four stages, shown as the steps under Becoming a hero in its Overview; the mission itself only ever shows its spark tile (D0850; see the hero world). Nothing else about the mission changes: its members, its CORA.md, its connections and its chat stay as they are. A hero that is on is listed to everyone in the organisation, so any other mission can add it, from Heroes helping here in its Tools tab, and then call it in (see Calling in a hero). Adding needs no approval from the hero’s mission (D0719). What they see is only its hero card: its name, its @handle, a description and up to five skills, each with examples. They never see its documents, its chat or its tools (D0693). The Heroes page does show everyone the organisation connections it uses, as chips, and its own page names its admins and the mission it came from.

What you see.

  • The Heroes page (US-01M7Q2NX, D0856). Each hero is a card: its mini hero on a band in its cape’s light colour, its name, where it came from (“Hero of the Page checks mission” or “Made as a hero”), what it is good at, its skills and connections as chips, “Helped N missions this week · N admins”, Open and its @handle. The filters are All, Mine and Off, each with its count; a hero that is off shows only to its admins and to org admins. + Create a hero is a button level with the page’s title, styled like Home’s + Create a mission (US-01M7SS78, D0881), and a dashed card after the heroes. The missions list has no Heroes filter or group any more.
  • A hero’s page (US-01M7Q2NX). The band with its mini hero, name, origin, @handle and state, Turn it off or Turn it on for its admins, and for the people in its mission a link to its mission or training room and four tabs: Overview (its look with Shuffle and Choose; its profile: Hero name with the @handle it gives, What is it good at?, When should Cora call it in?, How should it work? and Who can call it in?; its admins; and What it has done, which holds the card described below), Memory, Skills and Tools. Everyone else in the organisation reads its card part only. Saving the profile sends only what changed; a new name changes the @handle, and a name another hero’s @handle already has is refused: “Another hero in your organisation goes by @kit. Choose another name.” Who can call it in is every mission, or only missions its admins choose from those they are in; a mission not chosen cannot add it, and one that added it before shows it as “Not for this mission” and cannot call it.
  • Create a hero (US-01M7Q2DK, D0857). Hero name (with its @handle), the look (Shuffle, Choose), What is it good at?, When should Cora call it in?, How should it work?, Skills, Tools and connections (Web access; the organisation’s connections for an org admin), Hero admins (you, always ticked, and you can tick other members of the organisation as admins too) and Who can call it in?. Create hero also makes its training room, “’s training room”, and opens the hero’s page. If a chosen skill, connection or web access cannot be switched on there (an operator has blocked a connection, say), the form stays open with one line, such as “Scout was created, but these could not be switched on: the connection stock-feed. You can switch them on in its training room.”, and a link to open the hero. If someone else creates a hero of the same name at the same moment, yours is refused like any name already taken.
  • What it has done (the card; it was the Hero tab of mission details until US-01M7Q2DK; US-01M7MXD1, D0829; US-01M7P0GQ; US-01M7FGAJ). It is headed Make this mission a hero, with one line: “A hero is a mission that other missions can call on for help. Make one when this mission does a job well enough to share.” The hero’s look is not here: it is under Look on the same page, with Shuffle and Choose for its admins, and “Only the hero’s admins can change its look.” for everyone else (see the hero world). Then What being a hero means, in four lines: every mission in the organisation can call on this hero; they see only its card, never this mission’s chat or files; it works for them with everything this mission has (its skills, connections, vault keys and sandbox); and it can read this mission’s team folder but cannot change it while helping others. At the foot is the button Make it a hero for the mission’s admins, with “A mission admin can turn it back into a mission at any time.” There is no switch. Once it is a hero, the section is headed This mission is a hero, its line says “Other missions can call on this hero for help. They see only its card.”, and it shows the card (name, @handle, description, and each skill with its example prompts), “Card version N”, what callers get back, who switched it on, and for the mission’s admins Edit card and Turn it back into a mission; members see the card read only. The warning about personal connections shows before it becomes a hero when there are any (D0708). Making a mission a hero means anyone in the organisation who can call it can make it use everything it has; every Ask first step still comes to this mission’s people (D0794). A called hero reads the organisation’s team folder but cannot write to it; its working files go to a private folder that is emptied when its step ends (D0805). Below the card come Callers and spending and Incoming tasks.
  • The card editor. On the hero’s page it opens when you press Edit card, or Make it a hero on a hero that was turned back into a mission. It is headed What other missions will see, with Name, Handle (shown with @ in front; the API ignores what is typed there, because the handle always follows the name, D0856 (4)), What it is good at, up to five Skills (a name, and examples one per line), What callers get back (A short summary, the default, or A full answer), the button Make it a hero (Save card when the mission is already a hero) and Cancel. Before the editor opens for the first time, Cora drafts the card from the mission’s name, objectives and CORA.md (“Cora is drafting the card…”), and you check and edit it before you save. Focus goes back to the button that opened the editor when it closes.

What you can do.

  • Make a mission a hero. In the mission’s Overview, once its four steps are done, press Make it a hero. A dialog headed “Make a hero” asks for the Hero name, What is it good at? and When should Cora call it in?, says what stays the same, and has Make it a hero (Make it a hero anyway after the personal connections warning) and Not now. Cora does not draft this one: the draft is for the card editor on the hero’s page. The hero is on only once it is saved, so no hero is ever on with a card nobody checked. The panel shows who first switched it on. If two people save a new hero’s first card at the same moment, one is kept and the other sees “Someone else saved this hero card at the same moment. Reload the page and check it before you save again.”
  • Edit the card. Edit card opens the editor with the saved card. Every save raises the card version by one, so callers always see the latest card.
  • Switch a hero off. Press Turn it back into a mission: the hero is off at once and gone from the lists other missions see (the Heroes page still shows it to its admins and to org admins, under Off), and a line says “It is a mission again. Its card is kept for next time.” Later the section says “It was a hero before. Its card is kept for next time.”, and making it a hero again opens the kept card in the editor.
  • Make a new mission a hero. The New mission page no longer has a hero switch (US-01M7PG57, D0852): create the mission, or create a hero on the Heroes page. Once a mission’s four steps are done, its admins make it a hero from its Overview, with Make it a hero. Whoever creates a mission is its admin.

How a hero looks (US-01M7E5X2, US-01M7P0GQ). A hero that is on is easy to tell from an ordinary mission, wherever you see the mission:

  • Its name. The mission list and the mission header show the mission’s own name, with the chip described next; the hero’s name and @handle are on the Heroes page and on the hero’s own page. A hero that is switched off, or held off by an org admin, loses the chip and looks like an ordinary mission again.
  • Its chip. The mission keeps its spark tile and its own name; beside the name, in the list and in the header, a chip shows the hero’s pixel face and “Hero: ” in words, so it never depends on colour alone (D0850).
  • Its own page. Heroes are listed on the Heroes page, not in the mission list, which has no Heroes heading any more (US-01M7Q2NX, D0856). A hero’s mission is listed with the other missions, with its Hero chip.
  • Search. The search box finds a hero by its name or its @handle (with or without the @), and still finds it by the mission’s name.
  • Who sees it. Only members of the mission see its row and header, so this shows nothing new: the card fields (name, @handle) are what every member of the organisation can already read.
  • Small screens. At 390 pixels wide the name, handle and badge wrap inside the row and the header, with no sideways scrolling.

Who may (D0694, D0707). A mission’s lead is its admins. The mission’s admins (the person who created it and anyone they made an admin) promote it to a hero, edit its card, draft it and choose what callers get back. The Project Lead role does not make someone a mission’s lead, and being an org admin is not enough either: the API refuses an org admin who is in the mission but not one of its admins (403). Today the console still shows such an org admin Make it a hero in Overview, and the save is refused (BG-01M7RGV5). An org admin who is not in the mission cannot promote it or draft its card, because both read the mission’s CORA.md, which only its members can see. The mission’s admins or any org admin switch a hero off; an org admin can do it from Settings, Permissions without joining the mission, or from the hero’s own page when they are in the mission. Everyone else in the mission sees the section with no buttons and “Only the mission’s admins can change this.” The API refuses anyone else with 403, and anyone outside the organisation with 404.

Switched off by an org admin. When an org admin who is not one of the mission’s admins switches a hero off, it stays off: the mission’s admins see “Switched off by an org admin. An org admin must allow it before it can be switched on again.” and no Make it a hero button. Allowing it again does not switch it on: the mission’s admins make it a hero again by saving its card.

Where an org admin switches a hero off and allows it again (BG-01M7HDK9, D0728). There are two places, with the same rules (D0707):

  • Settings, Permissions, under Heroes, for every org admin, whether or not they are in the hero’s mission. It lists each hero that is on, with its card (name, @handle, description and what to ask) and a line such as “2 things to ask · in 1 mission”, and Switch off. Under Switched off by an org admin it lists the heroes held off, each with Allow again. After either, a line says what happened: “ is switched off.” or “ can be switched on again by the mission’s admins.” A refusal is shown in plain words. With no hero on it says “No hero is on. A mission’s admins can make it a hero from its mission details.” The block’s first line says what a hero is.
  • The hero’s page, under What it has done, for an org admin who is in its mission. One who is not one of the mission’s admins sees “As an org admin you can switch this hero off. It then stays off until an org admin allows it again.” with Switch off. On a hero an org admin switched off, every org admin there sees Allow again beside the “Switched off by an org admin” line; afterwards it says “ can be made a hero again by the mission’s admins.” An org admin who is also one of the mission’s admins turns it back with Turn it back into a mission, as its admin, and it is then not held off.

A member who is not an org admin is offered neither button anywhere.

Rules.

  • A handle is 2 to 32 lower-case letters, digits and hyphens, starting and ending with a letter or digit, and is unique in the organisation. It always follows the hero’s name (D0856 (4)), so a name whose handle another hero has is refused: “Another hero in your organisation goes by @kit. Choose another name.”
  • The draft is one model call, metered and counted towards the organisation’s AI usage. Unlike polishing a message, it stops when this month’s AI usage is used up and when the plan has ended, as Cora’s own work does: nobody is waiting on it to reply. It reads only the mission’s name, objectives and CORA.md: nothing from the mission’s documents, members, connections or tools goes into it. Nothing is stored until you save. When the AI usage is used up Cora does not draft: the editor opens empty with “Your organisation has used its AI usage for this month, or its plan has ended, so Cora did not draft the card. Write it yourself.” If drafting fails, the editor opens empty with “Cora could not draft the card just now. Write it yourself.”
  • In the Tools tab of mission details, Heroes helping here lists the heroes added to the mission, each drawn as its mini hero and named and described from its card (US-01M7MXD1, D0828; it took the place of the list the Connections tab had). With none, it says “No heroes are helping yet.”, and to anyone who is not one of the mission’s admins “No heroes are helping yet. A mission admin can add one.” The organisation’s whole directory of heroes is what Add a hero lists for a mission’s admins in the Tools tab, and what Settings, Permissions lists for org admins (D0719, BG-01M7HDK9).
  • Heroes only read in this release. A caller gets back what the hero’s card sets under What callers get back: a short summary (the default, up to 4,000 characters) or a full answer (up to 20,000) (D0692).

Callers and spending (US-01M7DRVS, US-01M7FGAJ; D0698, D0711, D0868, D0869). Under the card, a hero’s page has a Callers and spending section. Every member of the hero mission sees which missions have called the hero, and whether each one uses the default daily cap, has its own cap, or is stopped. Only the mission’s admins see what each mission spent today and this month, the caps themselves, and what each incoming task and run cost. They see these as a share of the organisation’s AI usage for the month (or for the whole free trial), never in dollars (BG-01M7QVN5). The console holds no credit figure: the API works the shares out. Everyone else reads “Each calling mission has a daily cap.” when a default is set, with no figures.

  • Default daily cap. The Daily cap for each calling mission: a percent of the organisation’s AI usage (for the month, or for the whole free trial) that applies to every mission that calls the hero (“no cap” when none is set). The mission’s admins see the percent, with Change, which opens a field with Save cap and Cancel; an empty field means no cap.
  • The table. One row for every mission that has called the hero: the mission and its Daily cap. Everyone but the mission’s admins reads Default, Own cap or Stopped. For the mission’s admins the row shows what the mission used on the hero Today and This month, as a share of the organisation’s AI usage (“3.7%”), and its cap: Default, the percent, or “0% (stopped)”. They can give a mission its own cap (Set, then a percent and Save) or take it away (Use the default). To stop a mission calling the hero, set its cap to 0.
  • What a cap may be (D0869). 0, which stops the calls, or a percent from 0.01 to 100 with no more than two decimals, such as 2.5 or 0.25. The cap is kept exactly as typed and reads back the same, whatever the organisation’s plan; it becomes an amount only when a call is checked against it, as that share of the organisation’s AI usage at the time, so it follows a change of plan. A percent above 0 but under 0.01 is refused, never saved as 0: “That cap is too small. Type 0.01 or more, or 0 to stop the calls.” More decimals are refused, not rounded: “Type a cap with no more than two decimals, such as 2.5 or 0.25.” A cap set before this rule keeps working as it was; an org admin among the mission’s admins sees it as a percent (“under 0.01%” when very small), and the other admins see Own cap (or “set” for the default) until someone types a new one. Nobody sees missions that have not called the hero. On a phone the table scrolls sideways inside its own box, never the page.
  • When a calling mission has reached its cap for the day, its call is refused before any AI work and Cora says so in its chat. The day starts again at midnight in the organisation’s time zone.

Incoming tasks (US-01M7DRDF, US-01M7FGAJ; D0697, D0711, D0868). The hero’s page also lists its recent Incoming tasks, newest first, for the hero mission’s members only, as compact rows: the state (Answered, Working, Waiting, Failed, Refused or Stopped), the calling mission, who asked, the first line of the question, how long it took and when. For the mission’s admins each row also shows its share of the organisation’s AI usage. Each row is a button that opens and closes it (aria-expanded) to show the full question, the answer rendered as safe Markdown (no raw HTML; only safe links; BG-01M7F9R0), or why there is none, and Show the run, which opens what the hero did for that task: its runs, how long each took, the tokens and the tools it used, and for the mission’s admins the share of AI usage. Show all tasks reads the next 20, and is hidden when there are no more. Anyone who is not a member of the hero mission gets “not found” for the run. Requests never appear in the hero mission’s chat, so they never enter the hero’s own work there. On the calling mission’s timeline, the Called in a hero card shows the question, the answer and how long it took, and no cost.

Personal connections in a hero (US-01M7DR9A, D0708). A hero uses every connection shared in its mission, personal ones included (see Personal connections).

  • The warning. When personal connections are shared in the mission, the Make it a hero dialog (or the card on the hero’s page) lists them with their owners before the mission becomes a hero: “Megan’s Salesforce, Henry’s Google Analytics 4: callers of this hero will be able to use these through the hero. They get the hero’s answer only, as a short summary or a full answer, whichever the hero’s card sets.” The card’s Make it a hero button then reads Make it a hero anyway. Without that confirmation the API refuses: “Confirm that callers of this hero may use the personal connections shared in this mission.” Saving a hero that is already on asks nothing. Organisation-wide connections are not listed.
  • The owner’s notice. Making a mission a hero posts one line in the mission for each owner of a personal connection shared there, other than the person making it a hero, mentioning them: “Your connection in can now be used by callers of hero @. Stop sharing it there if you don’t want that.”
  • The panel and the editor work at phone width (390 pixels) with no sideways scroll and use the console’s stylesheet only (no inline styles).

States. Not a hero yet. Drafting the card. The editor open (empty at the spend cap or when drafting failed). Saving, and a refusal in the API’s own words. On, with the card. Off, with the card kept. Switched off by an org admin. Greyed out for someone who may not change it.

Screens: Artifacts, reached from Artifacts on the left rail at /artifacts, and the Artifacts tab of mission details on a mission’s page (Documents and Shared in messages on one page). Until US-01M7PP0C they were Files (D0853). They are described in Artifacts and Documents. Cora’s memory is not here: the organisation’s is in Settings, under Cora’s memory, and a mission’s is the Memory tab (D0854).

What you see.

  • The organisation’s artifacts only (D0634, which replaced D0621’s Viewing picker; D0853; design sdlc-studio/designs/missions-and-heroes/OrgFiles.dc.html). A header row at the top holds the Artifacts title and the storage meter (US-01M71911), with the line “Documents Cora saved for the whole organisation.” under it. On the left, a search box, Search documents, then Documents with its count: each row “From Cora · ” and its date. With none yet it says “Nothing here yet. Cora writes documents here.” Under the list one quiet line says where a mission’s files are: “A mission’s files are in that mission: open it, then Mission details, Artifacts.” The page shows only the organisation’s documents (US-01M7SS78, D0883): the In missions list of D0853 is gone, so the page never lists missions, a mission’s files, the files shared in messages or Cora’s memory. /files?project=<id> and /projects/:projectId/files open that mission’s page with mission details open on its Artifacts tab (&doc=<id> opens a document in the chat’s place). /files?doc=<id> for a mission’s document goes there too (US-01M71911). On a phone ?files=open alone opens the details sheet.
  • The explorer on the left, like the one in VS Code. The edge between the tree and the viewer drags (220 to 520 px, arrow keys, double-click or Home resets) and is remembered per person. Small icon buttons on the Documents bar: Refresh (“Refresh documents”) and Collapse folders (D0567, D0621).
  • Artifacts stay current (US-01M704B7, D0557, D0558, D0621). A mission’s pane reads its Documents or Memory again within about half a second of a files live event, which the API publishes after Cora writes a document or a memory file and which carries only the section’s name, and reads both once when Cora finishes a run. The event goes only to the mission’s channel, so only its members are told (D0572, D0619), and nothing private travels with it (D0128). The organisation’s Artifacts page has no live channel: it reads again when the window regains focus or becomes visible and every 3 minutes while visible. Open folders, the selection and the open file are kept; a removed open file closes with a short note.
  • Folders and icons. A folder opens in place and shows what is inside it. Each folder has a caret that turns and a closed or open folder icon, each file has an icon for its type, rows are 26 px, and every level is indented by 16 px with a guide line. In a mission’s Memory tab CORA.md is pinned at the top, marked Always read (D0618).
  • The viewer on the right, filling the space (US-01M71911): text in a centred column of about 880 px with a Wide toggle (aria-pressed, remembered per person, never chat content, D0128), and for Markdown with three or more headings an On this page outline (h1 to h3, click to scroll, current section highlighted, Outline toggle, shown only where the viewer is about 960 px or wider). Markdown tables, in files and in chat, keep their heading case and scroll inside a bordered box. Markdown is formatted, HTML is shown in a locked-down frame with no scripts and no network, from a small page of the console with its own security headers so the document’s own styles apply (D0612). The frame never moves: the console removes redirects, forms that post somewhere and links that run code, and every link opens in a new browser tab (D0597), CSV is shown as a table, and PDFs and images are downloaded rather than shown (D0562). An SVG picture is shown through the same frame as HTML. An HTML or SVG file is shown only when it is self-contained; one that needs a script, a canvas or anything loaded from the internet is not framed, and every file that is not shown says Inline viewing is not supported for this file. Download it to view. with its download beside it (BG-01M7JAJE, D0736, D0737). An SVG document’s Download menu offers the original only. A PDF document Cora made shows a download panel instead of a preview: its name, its size and a Download button for the PDF (D0770). A picture Cora saved into Documents from her sandbox (PNG, JPEG, GIF or WebP) is shown inline, at most the pane’s width, with a Download button; one larger than 3 MB, or a CSV or PDF she saved, says Inline viewing is not supported for this file. Download it to view. with its Download button. A saved file downloads as it is only (US-01M7MKRM, D0807).
  • One header above the file: the breadcrumb on one line (the section, each folder, then the file), the name with its size and last-changed date, then Wide, Outline and Download.
  • The storage meter, a compact bar in the header row: how much of the team folder’s 5 GB is used, shown to every member as the amount used and the limit only.
  • On a mission’s page, the header’s panel button (at the right end of the header) opens and closes a pane on the right with that mission’s view. It is closed until you open it, and then remembers your choice. A file you choose opens in the middle in the chat’s place, with a bar (where it lives, its name, the download buttons) and a close button (or Esc); the message box stays. On a phone the pane is a full-screen sheet.

What you can do.

  • Read the organisation’s documents, and the documents of any mission you are a member of.
  • On a document: Ask Cora to change, which opens the mission’s chat with your message started; download it as PDF, Word (.docx) or the original; and go to its mission (D0569, D0570).
  • See a mission’s own artifacts in that mission: open it, then Mission details, Artifacts.

Rules.

  • Documents are written only by Cora. Cora’s memory is not shown here (D0854).
  • Every member sees the organisation’s documents. You see a mission’s documents only if you are a member of that mission (D0572, D0619).
  • Reference and Skills are not sections here: Reference files are in the organisation’s memory under reference/, shown as Reference files in Settings, Cora’s memory, and skills are built in. Each mission’s skills are listed and switched from the Skills tab of its mission details (D0618, D0620, D0630, D0631).
  • When the team folder is full, nothing more can be added to it until space is freed (D0531).

States. No documents yet (“Nothing here yet. Cora writes documents here.”). A search with no match. Team folder full. Document converting for download.

Screens: Connections is a page in the main area, like Missions, Heroes and Artifacts (US-01M7SS78, D0882; design sdlc-studio/designs/console-polish/proposal.html). From US-01M7GSCV (D0724) until then it was a window over the page you were on. The rail’s Connections button opens it at the address it always had, /connectors. There is one list, with what is connected first, and each item has its own page. It is for every member: the person sees what they can connect and what is connected; org admins also see the organisation’s setup (US-01M7AXT6, D0689). It is a re-arrangement of the earlier one-page list, which replaced the Tools screen. Nothing about the data, the rules, who may see or change what, or the connect and set-up flows changed.

The page.

  • No window. There is no dimmed area, no close button and no trap for Tab: move away with the rail like any other page.
  • Back to mission. A link from a mission (its Tools tab, the Connections section there, or one of Cora’s next-step buttons) opens the page with Back to mission at the top, which goes back to that mission (?mission=<id> in the address, kept while you move between the list and an item).
  • Old links and deep links land where they pointed. /connectors is the list, /connectors?view=connected (Manage connections until US-01M7SS78) opens the same list with focus on Connected, and /connectors?service=<key> (the address a service’s view always had) is that item’s page. ?show=<chip> chooses a chip, ?changes=tool-access shows the “Tool access is now set on each connection” card, and /connectors?tab=connections, ?show=mine and ?show=heroes, used by older links, open All. An old link to a hero’s page (/connectors?hero=<id>) does not open a page here (BG-01M7HDK9): if you are in the hero’s mission it opens that mission with mission details on Overview, which links to the hero’s page, even while the hero is switched off and however many missions you are in (BG-01M7HXJ2); if you are not, an org admin lands on Settings, Permissions (where the hero’s switch is) and anyone else on the Connections page. The chosen chip and the open item are kept in the address, never in the browser’s storage (D0128).
  • Focus. Opening an item’s page puts focus on its heading. Its back link puts focus on the button of the card that opened it.
  • Links into Settings. Settings, Permissions and Workspace usage open the Settings window, which closes back to Connections.
  • Unsaved text. If a dialog on the page holds typed text that is not saved, a click outside does not close that dialog: the rule is under “Closing a window by clicking outside it” in Settings (D0721).
  • Marks. Each service has a small tile with its initials. There are no third-party logos.

The list (/connectors). It has the heading “Connections”, Add any MCP server at the top right, and one plain line: “The services Cora can use in . Connect one, then choose the missions it is shared with. Anything that changes something waits for an approval.” Then the search box and the chips, then the groups below in this order. The cards fill the width: three per row on a wide screen (1,440 pixels and up), two on a laptop, one on a phone.

  • The search box, “Search connections”, narrows every group to the services whose name or owner matches. It does not find heroes.
  • Add any MCP server: for a service not in the list (see Add any MCP server below).
  • The chips (US-01M7FG9B). All, Analytics and search, Advertising and CRM and content (the names come from the catalogue’s data). There is no Heroes chip any more (BG-01M7HDK9). They carry no counts any more. The Mine and Organisation chips are gone: each connected card still says Yours, the owner’s name, or Organisation. The chosen chip is kept in the address (?show=).
  • Needs attention: first, and shown only when a connection is failing or needs a reconnect. A connector failing its check has Fix, for the person who can fix it; a connection whose access ended says “Needs reconnect”, with the service’s hint, and has Reconnect.
  • Connected (N): every organisation connection, your own connections, and other people’s connections shared in missions you can see. Each card has a mark, the name, an owner label (Yours, the owner’s name, or Organisation), a status and a one-line summary (for example the site, how many tools are on, who approves, which missions it is shared in). Your own and, for an org admin, the organisation’s have Manage; anything else has View. A connection that works but is not shared in a mission yet says “Not shared yet” and its button is Share in a mission. A member sees only what the mission pages already show them: the organisation’s connectors and the people’s connections shared in their missions (D0131). N counts your connections, the organisation’s and those shared in your missions. A hero is not a connection and is not counted (BG-01M7HDK9). With nothing connected the group says “Nothing is connected yet.” The “N connected” button that opened a second page is gone.
  • Ready to connect: the catalogue’s services you can connect now. Each card has the service’s mark, its name, what it can do (“Read only”, “Reporting only”) and how it connects (“Your Google key”, “Sign in with TikTok”, “Admin sets up once, then you sign in”), with Connect, the strong blue button. The names and the how-it-connects lines come from Cora’s catalogue data, not from the page.
  • Needs an admin to set up first: the services waiting for their organisation setup. An org admin sees “Organisation setup needed” with Set up, a plain button. Anyone else sees “Waiting for an org admin” with Ask an admin, which opens the service’s page with the names of the organisation’s admins to ask; only an org admin can set a service up. An organisation connection whose state a member cannot read is here too, with Ask an admin.
  • Not available yet: the services the operator has not set up, each with a “Not available yet” badge and no button. The group shows only when there is such a service.
  • What is not listed (BG-01M7HDK9). Web search and the code sandbox are part of Cora and are not listed as connections. Web access is set in Settings, Web access (see Web access page below), and the team folder’s size is on Workspace usage. Heroes are not listed either: see Hero.
  • When nothing is left to add it says “Every service in the list is connected. Add any MCP server for one that is not listed.” When nothing matches a search or a chip, it says so.
  • At the foot, together on one quiet line: a link to the guide Connecting an MCP server, a link to Workspace usage for org admins, and one line saying whether MCP is on (with a link to Settings, Permissions for org admins; anyone else reads that only an org admin can turn it on or off).
  • Organisation controls are in Settings, Permissions (US-01M7FG2A). It holds the Use MCP servers switch and one line: when on, “On. Cora can use the organisation’s connections. Turning this off stops Cora using every connection, without removing them.”; when off, “Off. Cora can’t use any connection; your settings are kept.” (D0315). The switch writes the policy from the version the page read, as before.

Add any MCP server. The button opens a small dialog headed “Add any MCP server”, for a service not listed, with the same two choices as before. Just for me connects with the service’s own sign-in (see Connect with the service’s sign-in below); it says “Not available yet” until the operator has turned it on. For everyone in is shown to org admins and opens the form for any other MCP server (see Any other MCP server below); it is off while MCP is off. Escape, Close or a click outside closes the dialog alone and returns focus to Add any MCP server.

What changed when access moved to the connection. The one-time notice about tool access opens the Connections page with a card, “Tool access is now set on each connection”, listing the missions that gained tools and which, and any personal connection whose sharing was stopped because none of the tools it allowed there still exist. Missions are named, never identified by their ids (D0716 (5), D0720 (6)).

One item’s page (?service=<key>). A back link at the top, Connections, returns to the list. Then a header: the mark, the name (the page’s heading), tags (Organisation, Yours or the owner’s name, and Read only where true), one line on what it does, and a link to its guide. Under the header are headed sections in place of the earlier numbered steps. The actions stay with the connection they act on, inside Account, because a service’s page can hold more than one connection.

A service each person connects (Google Analytics 4, Google Search Console, TikTok Ads, Salesforce, Meta Ads):

  • Organisation setup, first, only where the service needs one. An org admin sees the setup card and its dialogs, unchanged (see The setup cards below). A member sees Waiting for an org admin, the names of the organisation’s admins to ask, and no form; once it is set up the section says so.
  • Account. With no connection yet, one line on how it connects and a button: Add your key (for example Add your Google Analytics 4 key) for a Google key, or Connect (for example Connect TikTok Ads) for a service’s own sign-in (off until the organisation setup is done, and while MCP is off). With one, the person’s card: its status (Working, Connected, Failing with the reason, Needs reconnect with the service’s hint), “Service account ” or Connected as , Read-only, “N of M tools on”, and its own actions: Replace key or Reconnect, Check again, Remove or Disconnect, and for an org admin’s own key connection Make available to the whole organisation.
  • Tool access (US-01M7FG9B, D0716), once there is a connection:
    • the line “What Cora may do with this connection, set by its owner. Each mission only switches the connection on or off.” (for an organisation connection: “What Cora may do with this connection, set once for the whole organisation. Each mission only switches the connection on or off.”);
    • group settings: each group of its tools with its name, what it covers and its tool count, and one Allowed / Ask first / Off choice that saves at once; a group whose tools differ shows Mixed; Allowed is not offered for a group holding a tool that always asks. The group names are the catalogue’s own, or Read and Make changes. Show each of the N tools opens a choice per tool. Someone who may not set access sees the values read-only with “Only organisation admins can change access. For a personal connection, its owner sets it.”
  • Shared with missions. The missions it is shared in, a chip for each with a × that stops sharing, and + Share in a mission (a menu of the person’s missions). Only the owner shares (D0643); a mission has at most one connection of each kind (D0644). Each tool’s access is set once under Tool access by the connection’s owner and is the same in every mission it is shared in (D0716). Sharing into a mission whose hero is on first opens Share in a hero mission with the same warning as the Hero panel (“ is the hero @: callers of this hero will be able to use your through the hero. They get the hero’s answer only, as a short summary or a full answer, whichever the hero’s card sets.”), Share anyway and Cancel; focus goes back to the menu. Under each of the owner’s own connections, Hero calls that used this connection lists each hero call that used it: the calling mission, the person who asked and when, newest first (“No hero call has used it yet.” when none), for the owner only (US-01M7DR9A, D0708).
  • Everyone’s connections, for an org admin: each person’s connection for that service with its owner, status and missions, with Stop sharing on each mission and Remove. There is no key, no Test and no Use: an admin can only look, stop sharing or remove (D0649). A member sees other people’s connections they can see as one line each: “’s connection, shared in . Only can change it.”
  • Information: What it can do, How it connects, Who can use it and Category (the catalogue’s type: Analytics and search, Advertising or CRM, content and documents).

An organisation connection (Umbraco CMS, connected once for everyone):

  • Account: the setup card (see The setup cards below), for an org admin. A member sees Waiting for an org admin and who to ask, or that an org admin has set it up.
  • Tool access: for an org admin while MCP is on, the connector’s card (see An organisation connector’s card below) and the access groups, as above. Otherwise one line: org admins choose which of its tools Cora may use, and every draft tool always waits for an approval.
  • Approvals, for an org admin: the role that approves anything that changes something. Change it on the card under Account.
  • Used in missions: for an org admin, Used in N missions with their names. Everyone reads “It is one connection for everyone. Mission members switch it on in each mission’s details, under Connections.”
  • Information: What it can do, Who can use it and Category.

An MCP server an org admin added for everyone:

  • Account: its card, with Refresh and the menu Edit and Replace secret (see An organisation connector’s card below). While MCP is off the card is hidden and the section says so; its settings are kept.
  • Tool access: the access groups, as above.
  • Approvals: the role that approves, with a Change button that opens Edit.
  • Used in missions: as for an organisation connection.
  • Information: the server’s address and who can use it.

A member who opens it sees its name, the Organisation tag, that only org admins change it, and the missions it can be used in.

The setup cards (in Organisation setup for Salesforce and Meta Ads, and in Account for Umbraco CMS; org admins only, unchanged):

  • Salesforce, Meta Ads, Gmail and Outlook (the organisation’s own sign-in app; US-01M77WRF, D0682; Gmail and Outlook US-01M7KFB3, D0801). The card shows the callback address to give the service’s admin (<console address>/v1/oauth/callback) with Copy, the service admin’s steps and a guide link, and either Not set up with Set up, or the My Domain address, the environment (Production, Sandbox or Developer Edition), the consumer key, Secret set and who set it, with Change, Replace secret and Remove. Salesforce’s dialog asks for the My Domain address, the Consumer key and the Consumer secret (a password field, empty after every save; the secret is never shown again). Change with another key or address and Remove say first that everyone’s connection will need to reconnect. See Connecting Salesforce. Meta Ads (US-01M7943S, D0688) is a public app with one sign-in for everyone: its dialog asks only for the App ID (digits only; no address and no secret field), and once set up the card shows the server address mcp.facebook.com/ads, the App ID and No secret needed. Set up with who set it, and offers Change and Remove but no Replace secret. See Connecting Meta Ads.
  • Umbraco CMS (an organisation connection, connected once for everyone; US-01M78ND4, D0683, D0684, D0687). The card shows one line on what it does and a guide link. When the operator has not set the hosted server up yet it says it is not ready yet and has no Set up. When it is not set up: Set up. When it is: the site address, the client id, Secret set , Approvals: , the status (Working, Failing with the reason, Off or Blocked by the operator), “N of M tools on” and how many missions use it, with Test connection (“Connected. Cora can see 12 top-level pages.”), Change, Replace secret and Remove (asks first: it stops Cora using the site in every mission). The dialog asks for the Site address (https, just the site), the Client id, the Client secret (a password field, empty after every save; needed for a first set-up and whenever the address or client id changes; never shown again) and Who approves drafts (admin roles only), and has its own Test connection for the typed values. See Connecting Umbraco.

Google Analytics 4 and Google Search Console (US-01M746ZB, US-01M77R3E; your Google key). Add your key opens a dialog, which ends with Add connection, with the address as “Hosted by Cora”, the sign-in fixed to Google service account, the key file input (read in the browser and sent once; the page then shows only the account’s email), the steps to add the email as Viewer under Admin, Property access management (Search Console: Settings, Users and permissions, Add user, Restricted is enough), and Test connection (“Connected. 3 properties found: …”, or that it can see none yet). Replace key uses the same file input; Remove asks first: the key is deleted at once and the connection leaves every mission. Make available to the whole organisation (org admins only, on their own key connection; D0647): a dialog explains that mission members then switch it on or off in each mission and that Cora uses the admin’s key for everyone, lists the missions it will stop being shared in, and (for a kind with write tools) asks for the approval role; it is then an Organisation connection. See Personal connections, Connecting Google Analytics and Connecting Google Search Console.

Connect with the service’s sign-in (US-01M743JS, D0652 to D0655, mockup sdlc-studio/designs/mcp-oauth-connect.html). Connect opens a dialog. For a catalogue service the short name and the server address are filled in; for Add any MCP server, Just for me, it shows a Short name and a Server address, and when the address is typed one line says who runs the sign-in and what it asks for, or why the server cannot be used; when a different domain runs the sign-in the line adds “This server signs in through . Only continue if you trust it.” Connect… opens the service’s page in a new window; the dialog says “Waiting for you to approve in the other window” and closes when the connection is saved. The answer comes back to the same browser (the page the person pressed Connect in keeps a secret in memory, and only it can finish the sign-in); the dialog says “This sign-in took too long. Start again.” after ten minutes. When access has ended the connection is under Needs attention, the first group of the Connections page, as Needs reconnect, its card says “Access ended at . Cora skips this connection until you reconnect.” and Reconnect reruns the flow and keeps the shares. There is no Replace key and no Make available to the whole organisation for this sign-in. See Connecting with the service’s sign-in.

  • TikTok Ads (US-01M77T3F, D0679, D0681; Connecting TikTok Ads): Sign in with TikTok. Cora’s TikTok tools are a fixed list of names that only read, kept by Cora and not by TikTok’s own labels; any other TikTok tool is never listed, switched on or given to Cora. When TikTok ends the sign-in (after about 30 days) the card says Needs reconnect and adds “TikTok ends each sign-in after about 30 days.”
  • Salesforce (US-01M77WRF, D0680, D0682; Connecting Salesforce): until an org admin has done the Organisation setup, Connect… stays off and the Account section says “Salesforce is not set up for your organisation yet.”; once set up, the short name and the organisation’s server address are filled in and fixed, and for 30 minutes after the set-up a line says Salesforce can take that long to accept a new app. If Salesforce’s page sends back that it does not know the app, the dialog says the app may still be starting. The card shows Read-only.
  • Gmail and Outlook (US-01M7KFB3, US-01M7KFD6, D0800, D0801; Connecting Gmail, Connecting Outlook): read your own mailbox through the mail server Cora hosts. They sit in the CRM, content and documents group and are listed only once the operator has deployed the mail server. An org admin first sets up the organisation’s app: for Gmail the Client ID (ending in .apps.googleusercontent.com) and Client secret, with no address; for Outlook the Directory (tenant) ID, the Application (client) ID (both GUIDs) and the Client secret value. Then each person chooses Connect… with the short name and address filled in and fixed, and signs in on Google’s or Microsoft’s page. Cora asks for gmail.readonly only, or for Microsoft Graph’s Mail.Read, User.Read and offline_access only. The card shows Read-only; the three tools (search messages, read a message, list labels or folders) are Allowed.
  • Meta Ads (US-01M7943S, D0685, D0688; Connecting Meta Ads): the same way, with https://mcp.facebook.com/ads filled in and fixed. Cora asks Meta for ads_mcp_management and ads_read only, and its Meta tools are a fixed list of reporting tools that only read. Meta gives no way to renew a sign-in, so when it ends (after a few hours) the card says Needs reconnect and adds “Meta ends each sign-in after a few hours and gives no way to renew it.”

An organisation connector’s card (US-01M71Q42; under Tool access on Umbraco CMS’s page and under Account on the page of each MCP server added for everyone, org admins only):

  • the short name in normal case, and the address under it in a small monospace line that stops with “…” and shows the whole address on hover;
  • small facts: how it signs in (No sign-in, Bearer token, Header name or the service’s API user, with “secret stored”), Approvals: , and “N of M tools on”;
  • a status badge: Working, Failing with the reason in words, Off or Blocked by the operator, and when it was last checked;
  • Tools (opens the list inside the card), Refresh (lists the server’s tools again) and a three-dot menu with Edit, Replace secret (for Umbraco the menu sends you to its setup card under Account, where its address, sign-in and secret change);
  • inside the card, each tool shows its name as Cora sees it, its description and one Allowed / Ask first / Off choice (US-01M7FGB9, D0716). Allowed runs without an approval card, Ask first waits for one on each call, Off is never used. Every tool starts Off; the server’s own read-only hint only shows (D0131). A tool the catalogue marks always-ask is never offered Allowed (D0684), and a tool that only reads by the catalogue is never offered Ask first. An org admin sets organisation connections; on a personal connection only its owner does, and everyone else sees the setting read-only. A tool that is gone or whose description changed says so and is Off until it is set again.

Any other MCP server (org admins, from Add any MCP server, For everyone in ): the dialog explains each field: short name, server address, sign-in (No sign-in, Bearer token or A named header), the header name when it is a header, the secret, and who approves its write tools (admin roles only). Test connection checks the address and the secret and says “Connected. N tools found.” or why it failed, before anything is saved. Add server saves it, lists its tools, all off, and opens its page. The secret goes straight to the secrets store and is never shown again (D0131).

Rules.

  • Test connection runs on the server through the same address checks and the same egress proxy as a refresh: public https addresses only, never a private or internal address, no redirects followed (D0131). It stores nothing and writes no audit entry. The secret is used for that one connection and is never logged, kept or sent back. It has its own limit of ten tries in a burst, then one every ten seconds, for each person.
  • Every tool set to Ask first waits for an approval on each call, from the role named on the connector (D0132). A change of setting writes one audit entry naming the tool or group, and the old and new setting.
  • With MCP off, nothing new can be connected, shared or added; the settings are kept and come back when MCP is turned on.
  • Replace secret opens a small dialog with one password field; the old secret is never shown. Edit changes the address, the sign-in and the approval role; the short name cannot change because tool names and approvals carry it.
  • A connector cannot be removed in the console yet: the API has no remove route, so there is no Remove item (a follow-up story). Turn its tools off instead. Personal connections can be removed.
  • A personal connection’s owner sets each tool’s access on the connection’s card (D0716, replacing D0657’s checkbox): a tool set to Ask first waits for an approval card sent to its owner, or to the approval role the owner named. The owner always sees Approve and Reject on a card sent to them, including a card made before this change. When access moved to the connection, a personal connection’s tools took their setting from the old checkbox: off, or a tool the server marks read-only, became Allowed; on, the rest became Ask first. An organisation-wide connection keeps the approval role rule (D0132).
  • When access moved to the connection, each mission that had allowed only some of a connection’s tools was switched to the connection’s setting, so it gained the rest; a mission whose list named no tool that still exists was switched off (a tool that was only switched off at the time still counted, so that mission stayed on). Each org admin got one notice listing the missions whose organisation connections gained tools, and each personal connection’s owner one for theirs, including any share that was stopped (D0716 (5)).
  • The page asks only the routes it may: a member never calls an org admin route, and sees organisation connectors and colleagues’ connections only through the missions they belong to. Nothing is stored in the browser.
  • The names and addresses are plain text in cards, never inside a table header, so they are not upper case.
  • The page works at phone width (390 pixels) as one column with no sideways scroll, and its main controls are at least 44 pixels tall. It uses the console’s stylesheet only (no inline styles, the console’s CSP) and returns focus to the button that opened a dialog or an item’s page.

What you see. The Web access part of Settings, at /web-access (US-01M71PTP, D0637). Org admins open it from the Settings menu; a member has no menu item for it but sees the card read-only at its address. The Connections page does not link to it. It shows one card, Web access, loaded as its own chunk so the console’s first load does not carry it. It has a switch “Let Cora use the web” (on for a new organisation; an organisation made before 17 November 2026 keeps its setting) with two options under it: “Search and read pages” (it comes with web access and has no switch of its own) and “Use a browser” (on by default; with it on, Cora can open pages in a real browser where the team sandbox is on, and the line says “Not set up for your organisation yet” until the organisation’s browser is ready; see Cora’s browser). While web access is off both options are greyed and unticked and say “Turn on web access first”, so a tick never reads as on. Then a list “Missions: On by default / Off by default”, a box “Blocked sites” (one per line; a pasted address is cleaned to its site name), a number “Monthly search limit” (0 to 100,000; each search uses a little of the organisation’s AI usage) with how many searches have run this month, and a line that every search and page address goes into the audit log and page contents never do. Save sends the version the page read; if someone else changed the settings first, the page reloads them and asks you to make your change again. An entry that is not a site name is refused with the reason. Members see the same card read-only, with no Save button. A mission admin turns web access off or on for one mission in that mission’s Details.

Screens: Workspace (/workspace), linked for org admins from the Connections page and from Settings, Usage and billing (Team workspace usage). It opens inside Settings, with Usage and billing marked.

What you see. The organisation’s team sandbox: its state, how much of the 5 GB is used, in GB with a bar, and when the folder was last measured.

Rules.

  • Only an org admin sees the usage and the links to the page. A member who opens /workspace sees the heading, “The workspace is where Cora runs commands and keeps working files for your organisation. Only an org admin can see how much of it is used.” and the guide. The usage route answers every member with the size against the limit only, for the Artifacts storage meter; its state and when it was measured are for org admins (D0531).
  • At 90% the page shows a warning; at 100% it says Cora cannot write files or install tools in the sandbox until space is freed (D0531).

States. Not available yet (“The team sandbox is not switched on yet. Cora works without it for now; this page shows the usage once it is.”), not yet set up, being set up, ready, setup failed. Under 90%, warning (90% or more), full (100%).

What you see. Settings in the account menu opens one window over the page you were on, at /settings/general (US-01M7FG2A, D0715 (5); mockup screen 4). On the left is a menu in two groups, and the chosen part shows on the right; its name is in the address, so a link can open it. ×, Escape and a click on the dimmed area outside the window close it, go back to the page it was opened from and put focus back on what opened it: the link on that page, or the account button when it was opened from the account menu. Opened by its address, it closes to Home. On a phone the window fills the screen. Opened at General it shows the menu as a list first; opened at any other part (People and roles from /members, say) it shows that part at once. Each part has Back to the list.

You

  • General: your name (change it and press Save name), your email, your organisation and your role in it, Switch organisation (only when you belong to more than one) and Log out (the same “Sign out of Cora?” check as the account menu). Send with Enter: a switch, on by default. On, Enter sends a message and Shift+Enter starts a new line. Off, Enter starts a new line and Ctrl+Enter (Cmd+Enter on a Mac) sends; the hint under the message box names the key your computer uses. The choice is kept in this browser for you only; it is a setting, never anything you wrote, and if the browser keeps nothing it reads as on.
  • My connections: a link to the Connected group of the Connections page, where your own connections are listed.

Organisation (org admins see every part; a member sees People and roles and Cora’s memory)

  • People and roles: two tabs, Members (the people in the organisation, the pending invites and Invite members; see Members page below) and Roles (the role catalogue; see Roles page below). A member reads both and can leave the organisation from their own card; only an org admin changes anything.
  • Permissions: the organisation’s controls: whether Cora may use MCP servers (moved here from the Connections screen), and Heroes, where an org admin switches a hero off and allows it again (BG-01M7HDK9; see Hero).
  • Cora’s memory (US-01M7PP0C, D0854; design sdlc-studio/designs/missions-and-heroes/Memory.dc.html) at /settings/memory: what Cora knows about your organisation, in every mission. Every member can open it. Three parts of the organisation’s memory/ folder: CORA.md, marked Always read, with the date it was last changed (memory/CORA.md); Reference files (“Facts Cora should rely on: price lists, product details, policies.”; memory/reference/), each with its size and date; and What Cora has learned (“Notes from earlier work that Cora reads when she needs them.”; everything else in memory, its folders opening in place). Choosing a file opens it on the same page, with an × button, “Close the file and go back to Cora’s memory”. Who may change it: the organisation’s admins and the platform super admin (D0618, D0855); every change is audited without its text. To them the page says “You can change it; members can read it.” and shows Edit on CORA.md (the whole text in a box, with Save and Cancel), Add a file under Reference files (one text file, .md, .markdown, .txt or .csv, up to 60,000 characters; the reference folder is made first if it is not there) and Remove on a file at the top level of Reference files or What Cora has learned (not on a file inside a folder), which asks first. Everyone else reads “Members can read it. Org admins can change it.” and sees no change controls. The service checks every change again. Until the team folder is ready it says “Your team folder is being set up. Try again in a few minutes.” An old link /files?path=memory/<path> opens that file here.
  • Web access: whether Cora may use the web (search, read pages, use a browser), and its blocked sites.
  • Usage and billing: the plan and billing, and the team workspace’s usage.

Where the old pages went. The old addresses keep working, so links in emails and in these docs still open the right place: /members and /roles open People and roles at its Members and Roles tabs, /web-access opens Web access, /workspace and /settings/billing open Usage and billing.

Closing a window by clicking outside it (US-01M7GSCV, D0721). Every window that dims the page behind it closes when you click the dimmed area: the Settings window and each dialog in the console (inviting a member, editing roles, the confirmations, adding a connection, adding people or a hero to a mission, and the rest). It is the same as pressing its Cancel or ×, and Escape still works. One safeguard: if you have typed text in the window that is not saved yet (an email address in the invite dialog, a new name in Settings), a click outside does nothing, so a slip of the hand never throws your typing away; use Cancel, × or Escape to close it and discard the text. A search box does not count as typing. Text counts as saved only once a save has worked: if a save is refused (a blocked site Cora does not accept, say), the window stays open with your text, and after a save that worked a click outside closes it (BG-01M7GZVF). The whole click must be on the dimmed area: a press that starts outside and ends inside the window, or the other way round, closes nothing. A dialog on the Connections page follows the same rules. A dialog opened inside Settings closes alone and leaves Settings open. The docked mission details pane is not a window of this kind and stays open.

Not in Settings yet. There is no theme, language, default model or notification setting: nothing behind them exists yet, so none is shown (D0715 (5)).

What you see. The members page is the Members tab of People and roles in the Settings window (US-01M7GSCV, D0722): open it from the account menu’s People and roles, or at /members. It shows one card for each person in the organisation, then a Pending invites section with one card for each invite waiting to be accepted. A person’s card shows an avatar with their initials, their name with their email under it (both in normal case, on one line each; the name is the part of the email before the @ when none is set), an Org admin or Member badge, their roles as chips (“No roles yet, so Cora can’t route work to them.” when they hold none) and one actions menu. Your own card says “You”. An invite’s card has a dashed envelope avatar, the address, an Expires badge (amber when it expires within 24 hours), the org role it starts with, its starting roles as chips, and, for org admins only, a Copy link button and an actions menu. At the top right, + Invite members (org admins only) opens the invite dialog, headed “Invite a member to” and the organisation’s name. Under the heading are a search box, a switch (All, Org admins, Pending, No roles) and “N people · M pending”. On a phone each card stacks: the name and menu, then the org role, then the roles.

What you can do. An org admin’s menu on a person has Edit roles (a dialog with a checkbox and a one-line description for every role that is not retired; Save sends only the changes), Make org admin or Make member (PATCH /v1/orgs/{orgId}/members/{userId}), and Remove from organisation…, which asks first. The only org admin has no Make member and no leave action; the menu says to make someone else an org admin first, and the API refuses the leave or removal (409, rule LAST_ORG_ADMIN, D0633). On your own card the last item is Leave organisation…, which asks first. A member sees only their own card’s menu, with Leave organisation…. An invite’s menu has Resend (new link, 7 more days) (POST /v1/orgs/{orgId}/invites/{inviteId}/resend: a new invite for the same address, org role and roles, a fresh expiry, the old link no longer accepting; Cognito’s RESEND mails the temporary password again and restarts its expiry when the person has not signed in yet, and when they have, no email goes and the page says so; the address having joined since is a 409 INVITEE_ALREADY_MEMBER, “That person has already joined.”) and Revoke invite…, which asks first (DELETE /v1/orgs/{orgId}/invites/{inviteId}). Copy link puts the invitation link on the clipboard and says so; if the browser refuses, the link appears in a selectable field under the status line, with Dismiss, and stays until dismissed or until a later copy works. The invite dialog has the email address, the starting roles with their descriptions (at least one), and Member or Org admin with a line explaining each; after Send invite it shows the link with Copy link, Invite another and Done. Menus work from the keyboard: arrow keys move, Esc closes and puts focus back on the button. Dialogs keep focus inside, close on Esc and give focus back. A line under the heading, read out by screen readers, says what just changed.

Rules. Org admins are the organisation’s role; mission admins are separate (D0623) and the page never changes them. Changing someone’s org role is for org admins (403 otherwise); a demotion, a removal or a leave that would leave no org admin is refused (409, rule LAST_ORG_ADMIN), also when two changes race. Removing the last holder of the fallback decision role is refused with the fix named (FALLBACK_ROLE_HOLDER_REQUIRED). Every refusal is shown in the API’s own words. Nothing from a private chat is shown or kept (D0114, D0128).

What you see. The Roles tab of People and roles in Settings (at /roles) lists the organisation’s roles, one card for each role. Left to right a card shows the role (its title in normal case, its key in small grey monospace, an Admin role badge in blue and a Fallback decision role badge in amber, then its description as plain text), the people who hold it as chips, and, for an org admin, one actions button (three dots), named “Actions for ”. A description is cut to two lines with Show more and Show less; a role with no description says so, and an admin sees a hint to add one. On a phone the card stacks: the role and its menu on top, the holders below, and the page never scrolls sideways.

What you can do. A holder chip shows initials and a name, with the email in a tooltip. An org admin sees an x on each chip (named “Remove from ”) that removes the person at once and says so; + Assign opens a list of the organisation’s current members who do not hold the role, with a search box; pick one to assign. The actions menu has Edit role (a dialog for the title and description together; the older Edit description button is now part of it), Make admin role or Clear admin flag, and Retire role…, which opens a confirmation. The fallback decision role has neither Retire role nor Clear admin flag, because the API refuses both; the console cannot change which role is the fallback. The menu works from the keyboard: arrow keys move, Esc closes it and puts focus back on its button. + Add role is the button at the top right; its dialog, headed “Add a role”, has the title, the description with a live “N / 500” count and a hint, and an Admin role checkbox that says what it does. Dialogs keep focus inside and give it back when they close.

Search and filters. The search box matches role titles, descriptions and holder names. All, Admin roles and No holders narrow the list, and “N of M roles” says how many show.

Rules. The API decides who may change roles (org admins only) and refuses to leave the fallback decision role without a holder; the page then shows the refusal and the fix. A member sees the cards read only: no x, no + Assign, no menu and no + Add role. Retired roles stay listed with a Retired badge and no controls.

What you see. An organisation admin opens Billing from the account menu’s Billing and plan row (US-01M74YEN, D0668; the row does not open Billing for a member, and the address shows a member “Only an organisation admin can see billing. Ask one if the plan needs changing.”). The page has these parts:

  • Plan: the plan, monthly or yearly, and its state (Active, Payment failed, Ended, Free trial, Paused); Members (“11 of 15” on Pro; on Standard “7 (5 included, 2 extra at US$9 a month each)”); Next charge (“11 Dec 2026 · US$67”, or “None: the plan ends on 11 Dec 2026”). Change plan, Switch to yearly (or Switch to monthly) and Cancel plan (or Resume plan).
  • Choose a plan, shown when there is no paid plan: Standard or Pro, Monthly or Yearly (2 months free), Continue to payment and Talk to us about Enterprise.
  • AI usage this month: “Cora AI’s models: 30% of this month’s included AI usage used”, a meter, and when Cora pauses. Only Cora AI’s own models are shown, and never in dollars (BG-01M7QVN5); the plan’s price and the next charge stay in US dollars.
  • Payment method and invoices: “Visa ending 4242 · expires 08/2028”, Update in Stripe and All invoices in Stripe.
  • Billing details: the billing email and the ABN, from Stripe, and Update in Stripe.

Words. Back from Stripe: “Thank you. We are confirming your payment with Stripe.” and then “Thank you. Your Pro plan is active and Cora is back at work.” If Checkout was closed: “Checkout was cancelled. Nothing was charged.” Cancel asks: “Your plan keeps working until the end of the period you have paid for, 11 Dec 2026. After that Cora pauses and nothing is deleted.” with Cancel at the end of the period and Keep my plan.

Read from. GET /v1/orgs/{orgId}/billing (org admins; a member is 403), POST .../billing/checkout, .../portal, .../change and .../cancel.

What you see. There is no trial bar (D0717). While an organisation is on its free trial (US-01M74YKK, D0658, D0659), the trial shows in the account menu’s Billing and plan row (see The left rail above): Free trial, 9 days left, a meter with 34% of your trial’s AI usage used and Choose a plan for an organisation admin (it opens Settings, Billing; a member sees no link, because members cannot pay). One day left reads “1 day left”. An organisation admin sees the trial row throughout the trial. A member sees it only in the last 3 days, or when the trial is paused, and without the usage figure or the meter; before that a member sees the plan’s name. Nobody sees it once the organisation is on a paid plan, and it is not shown to an organisation made before the trial existed. If the plan cannot be read, the row is left out. So the trial is still noticed, the account button carries a dot when the trial has 3 days or less left, its usage is at 80% or more (org admins only) or it is paused.

When the trial is paused (its 14 days are over, or its AI usage is used up) the trial row reads Free trial ended or Free trial used up, with “Cora has paused her work”, and a screen opens once for each time the console loads: Your free trial has ended or Your free trial’s AI usage is used up. It says Cora has paused her work in all missions, the team can still read missions, files and messages, and nothing is lost. An admin sees Pro (15 members, 5x Standard’s AI usage, US$199 a month) and Standard (5 members, Standard AI usage, US$49 a month), a choice of plan (Pro first, Monthly or Yearly (2 months free)), Continue to payment (it goes to Stripe Checkout for the plan ticked), Compare plans (the pricing page) and Talk to us (an email to the Cora AI team, with the subject “Talk to us about Cora AI”), and Close. A member sees “Ask an organisation admin to choose a plan.”, Compare plans, Talk to us and Close. The screen can always be closed with Close or Escape, and focus goes back to where it was. Nothing is kept in the browser.

The limits. A trial refuses a 6th member invite and a 4th active mission, with the messages in State messages below. Home’s usage tile shows the percent of the trial’s AI usage used.

Read from. GET /v1/orgs/{orgId}/plan: any member reads the plan, its status, days left and limits, and on a paid plan the interval, the end of the period, whether it ends there and the date a failed payment’s grace ends; an org admin also gets the usage and the cap.

On a paid plan a notice at the top of the page shows only what needs attention (US-01M74YEN, D0665). After a failed payment an org admin sees Your last payment did not go through. Cora keeps working until 15 Dec 2026; update your card in Stripe to keep her going. with Update in Stripe; a member sees “The last payment for this plan did not go through. Ask an organisation admin to update the card.” When a plan has ended an org admin sees Your plan has ended and Cora has paused her work. Your team can still read and chat. with Choose a plan, and a member is told to ask an admin. A plan that runs well shows nothing.

The plans (D0656, D0860). Every plan has every feature. They differ only in members and AI usage. Prices are in US dollars, and paying yearly gives 2 months free. The figures come from the plan settings in packages/plans, listed on plan and trial variables.

Plan Price Members AI usage Missions
Free trial Free for 14 days, no card Up to 5 A quarter of Standard’s monthly AI usage, for the whole trial Up to 3 active
Standard US$49 a month 5 included, then US$9 a month for each extra member Standard AI usage each month No limit
Pro US$199 a month Up to 15 5x Standard’s AI usage each month No limit
Enterprise Talk to us Talk to us Talk to us No limit

AI usage is a share, never dollars (D0859 (6), D0868, BG-01M7QVN5). Wherever the console shows AI usage, it is a percent of what the plan includes: for the month on a paid plan, or for the whole free trial.

  • The account menu’s Billing and plan row, for org admins: “Pro · 42% used”, or during the trial a meter with 34% of your trial’s AI usage used (see The left rail).
  • Home’s AI usage this month tile (Trial AI usage during the trial), for org admins only (see Home).
  • Billing’s AI usage this month: “Cora AI’s models: 30% of this month’s included AI usage used” (see Settings, Billing).
  • A hero’s Callers and spending and Incoming tasks, for the hero mission’s admins only: what each calling mission used Today and This month, its cap, and each task’s share. Everyone else sees Default, Own cap or Stopped, and “Each calling mission has a daily cap.” when a default is set, with no figures (see Hero).

Token counts are not a price, so every mission member sees them: the mission’s “ tokens so far” under Usage in Overview, and “Thought for 1m 20s · 64.5k tokens” under How I worked on a reply. The Called in a hero card shows no cost. The only dollar amounts in the console are a plan’s price, the price of an extra member and the next charge, on Billing, in the plan choice and on the paused screen.

When the AI usage is used up. Cora pauses her work in every mission at her next step, and nothing is deleted. People can still read the missions, the files and the messages, and write in the mission chat. On a paid plan Cora starts again when the next month begins. On the free trial the trial is paused and the paused screen opens (see Free trial and paused screen). Drafting a hero card stops at the cap too (see Hero). What uses AI and what keeps working is on managing usage.

Who sees what. Only org admins see the organisation’s usage figures and Billing. A hero mission’s admins also see that hero’s spend and caps, as shares (see Hero). A member sees the plan’s name in the account menu, and the trial row only in the last 3 days of the trial or once it is paused, without the figure.

These are the exact words the console uses, each shown in the place it applies. A value in angle brackets, or a date or number in an example, is the live value.

Trial: too many members or missions (US-01M74YKK)

The free trial includes up to 5 members. Choose a plan to add more people.

The free trial includes up to 3 active missions. Choose a plan to start more.

Trial paused (the paused screen’s heading, one of the two, and its text)

Your free trial has ended

Your free trial’s AI usage is used up

Cora has paused her work in all missions. Your team can still read missions, files and messages. Choose a plan to carry on; nothing is lost.

A member sees, in place of the plan choice:

Ask an organisation admin to choose a plan.

Payment failed (the notice at the top of the page)

Your last payment did not go through. Cora keeps working until 15 Dec 2026; update your card in Stripe to keep her going.

That is for an org admin; the date is when the grace period ends. A member sees:

The last payment for this plan did not go through. Ask an organisation admin to update the card.

Plan ended

Your plan has ended and Cora has paused her work. Your team can still read and chat.

That is for an org admin, with Choose a plan. A member sees:

This organisation’s plan has ended and Cora has paused her work. Ask an organisation admin to choose a plan.

Billing, for a member

Only an organisation admin can see billing. Ask one if the plan needs changing.

Invitation sent (the invite dialog, after Send invite)

will see the invitation when they sign in with that address. You can also send them this link. It stays open for 7 days.

Invitation not for you (Accept your invitation)

This invitation is not for the address you are signed in with, or it has expired, or it has already been used. If you have already accepted it, you are a member: open your missions from the home page. Otherwise, sign in with the address the invitation was sent to, or ask the person who invited you for a new invitation.

Forgot password

If there is an account for that email, we have sent it a code. Enter the code and your new password.

It is the same whether or not an account exists for that address, so it never reveals who has one.