Omniscio

Mission Control — workspaces, people and publishing (part 3)

Part 3 of the Mission Control page: workspace membership and accounts, sharing a board or inviting someone into a workspace, the permissions and trust controls around them, and the strategic layer above a board — goals, portfolios, and publishing work to a public web address.

On this page

What it is

This is part 3 of the Mission Control page. It covers who can see and change a board, and the structures that sit above one: workspaces and their people, how a board or a workspace is shared, the permissions and trust controls around that, and the strategic layer of goals, portfolios and public publishing.

Where to find it

Inside Mission Control: workspace and member controls live in the workspace's own settings, the sharing actions on a board's menu, and goals and portfolios in the left navigation.

How it behaves

Workspaces

Workspaces follow the familiar workspace model across three surfaces.

The left-nav header shows a coloured icon tile (initial plus a home glyph) with the workspace name and a switcher popover. Its ⋯ actions menu offers Manage workspace, Edit workspace (icon colour, name and description), Sort workspace (a per-workspace display order), Add new workspace, Browse all workspaces, View archive/trash, and a confirm-gated Delete that is disabled for the Main workspace and for the last one.

The Manage-workspace page (/workspace/:id) has a banner with a large icon, the name, an inline-editable description, Feedback and Members controls under Recents · Content · Permissions tabs. Content is a boards-and-dashboards asset table with the creator avatar, creation date, last modified, folder, search and a Filter menu (Type, Created by, Folder including No folder, and Last modified: last 7 days, 8 to 30 days ago, over 30 days ago; ticks in one group match any of them, ticks in different groups must all match). It shows a "Showing N of M" count while anything narrows the list, an honest "No assets match your search or filters" message with a Clear button, selection-delete that only ever removes rows you can see, and a Cleanup mode that flags assets unmodified for 30+ days. Recents reuses Home's recent visits and favourite stars. Permissions sets the workspace access level (Open or Closed), manages workspace owners (people only; the last owner is guarded), and — for a Closed workspace — the specific members and teams with access, through a searchable picker over the account directory and teams. An Open workspace simply reads "everyone in your account".

The user-management page (/users) is the account directory as a table: an editable role dropdown, a Products multi-select (seeded with work management), status, Teams chips, joined and invited dates, an Invite dialog that reuses the member invite, and a Manage teams modal over a real Teams subsystem.

Archiving a workspace is separate from its access level. Archive sets the workspace's own archived flag and moves it to the switcher's Archived (N) list; Restore clears that flag. Neither touches the access level, so a Closed workspace stays Closed through an archive and a restore, and choosing Closed in Permissions never archives anything. Workspaces archived before this fix were stored as Closed instead, so they show in the active list as Closed workspaces until they are archived again.

The backend carries a long tail of schema migrations behind these features; none of it is exposed to the app, which talks to it over the documented REST routes.

The left sidebar shows only the sections you use. Each category (All Projects, Dashboards, Goals, Portfolios, Documents, Spreadsheets, Forms) has its own switch, in the Sidebar card of Settings → Mission Control and on the Sidebar card of Mission Control's own Settings page; both change the same setting, and an open sidebar follows a flip straight away. The Omniscio Settings card also shows "N of 7 shown" and a Show all checkbox that turns every section on or off at once (a dash means only some are on).

  • New users start lean: only All Projects shows. Creating a dashboard, goal, portfolio, document, spreadsheet or form switches its category on.
  • Upgrading never hides what you use: the first time the sidebar finishes loading with no error, every category that already holds something is switched on, once. It reads the workspace that is open, so a category used only in another workspace comes on when you create something there or flip its switch.
  • A choice you make later sticks: the one-time pass never runs again, so a category you switch off stays off.
  • Without the Omniscio host (browser or phone embed) every section shows.
  • One button adds a project: the + beside All Projects opens the New project picker, which offers Start from a template, Start from scratch and Import a project. The separate template and import icons are gone, so + is the only place to start one.

Account & passwords

The login screen has a "Forgot password?" link: enter your email and you get a neutral confirmation ("if that email exists, we've sent a reset link") that never reveals whether the address is registered. Because the plugin runs on an in-memory router, the reset itself happens on a backend-served HTML page the emailed link points to. The reset token is single-use, expires in one hour, and a fresh request supersedes any earlier link; completing a reset signs out all existing sessions.

Signed in, the top-right account menu adds "Change password" — a modal that requires your current password and a new one of at least 8 characters. Changing it signs out all your OTHER devices while the current one stays signed in.

Logout is server-side — signing out revokes the session on the backend too, so it takes effect even after a long idle. There is still no email verification on signup; accounts are usable immediately.

Your boards come back the moment you sign back in. When Omniscio signs you out, it clears this computer's copy of your Mission Control data. Signing in again pulls your projects back right away, with no restart needed; a routine sign-in renewal never re-downloads anything.

Board sharing & private-board enforcement

The board header has a Share button opening a modal with a copyable join link and the board's member list. The join link points at a backend-served join page; another user on the same account signs in there and is added to the board as an editor. Private boards are enforced end-to-end: only a board's members — owner, editor or viewer principals, held directly or via a team — can see a board in listings or open it. Only the board owner can mint, rotate or revoke the share link, or add and remove members.

Local board sharing (push-then-share). Boards created through the local database when the backend is unreachable are marked as pending sync. Clicking Share on one pushes it to the backend first, before the share UI appears. The push sends the board's local id as its client id and the backend stores the board under that same id, so links and saved references keep working after it syncs; only if that id is already taken does the backend assign a new one, and then every reference is moved to it atomically and the modal uses the resolved id. The same guard runs on the CLI share routes. If the push fails, a clear error is shown and sharing is disabled until connectivity returns.

A team guest is capped below its board role

The team role guest is a ceiling on whatever board role a guest was given. A guest sees only the boards they were added to, never the rest of the account's boards, and on those boards they read and comment and nothing else: no value edits, no items, no columns, no group renames, no board settings, no archive or delete, no board creation, no invites. The cap holds even when a guest was given a board owner or editor row by mistake. An account viewer who holds an owner row is likewise read-only — and read-only rather than read-and-comment, so only a guest on a board they were added to keeps commenting.

One rule, on the server. A single capability function decides what a person may do with a board; the write, comment, archive/delete and create checks all delegate to it, and board listings filter a guest to their own boards at every board kind, not just private ones. The desktop mirrors the server's answer rather than re-deriving it: the server publishes "can write content" and "can comment" with each board, the sync stores them as nullable columns, and one desktop module reads them for the write guard and the add-comment handler. A board the server has not answered about yet keeps the plain role rule, and the columns start empty for exactly that reason, so an existing install is never locked out of its own boards before its first sync.

What a guest may never do. Create a board, a workspace, a folder or a dashboard. That one rule is checked by the folder, dashboard, workspace and template deploy routes before they build anything, and the desktop's own create doors apply it too — seven local create routes (template deploy-config, boards, documents, spreadsheets, forms, goals and portfolios) and the five IPC create channels behind them. A solo local install with nobody signed in has no role to check and can still create; a signed-in account whose role could not be read is refused, because an unread role is not a permission, and the refusal is logged rather than silent.

Facts the code used to carry only as comments, kept here. "May you see this board" and "may you write to it" are different questions, and a write route has to opt in explicitly; an account admin has two sources and both are needed (the account role column and an owner/admin principal on the account's main workspace); a private board is visible only to its principals, held directly or through a team, while a main or shareable board stays visible account-wide; restoring a deleted board authorises against the role held at deletion time, because the delete cascade removes the live member rows; a board whose workspace has been deleted is treated as gone; a board inside a closed workspace additionally requires workspace membership on top of anything the board itself says; and the refusal rule lives in one place so the CLI door and the desktop write guard cannot disagree.

Column hiding is a separate rule. Two options exist. "Hide from this view" (column menu) only tidies one view and hides nothing from anyone. "Keep private…" (column menu, level "Private") is real privacy: a column private from you is left out of the board export, search (a private column's text never makes an item match, and the assignee and status filters ignore it), the CLI board and item reads, and AI answers, and a published board or dashboard leaves out every column that is private from anyone. These rules live only on the desktop that set them for now; the phone, the web and other people's desktops do not see them yet.

The workspace ⋯ menu has "Copy invite link." It reuses the workspace's existing invite link and only mints a new one when none exists yet, so the link stays stable across copies. Only workspace owners can manage the link.

Accepting an invitation pulls the account's boards. Accepting adds you to the account server-side, but the app holds a local database of boards, so a join that only refreshed the page left a new member with none of the team's boards. The invite-accept screen now asks the main process to pull the account's workspaces and their boards before it reloads. It is best-effort and bounded — the wait is capped by the bridge's write deadline, so a slow or failed pull never blocks the join itself; the member simply joins without boards on that visit. The same pull runs once when Mission Control sync starts for a signed-in user, so a login or account switch discovers boards too. Each run leaves one summary line (workspaces found, boards synced, boards failed), which is what distinguishes "the pull found nothing" from "the pull did not run".

Removal takes the team's boards off the member's computer. When a member loses access to a workspace, the backend starts answering "not found" for that workspace's boards. The next pull hides that workspace and every board in it the backend now refuses. It only hides them and never erases them: attachment files stay on the computer, nothing on the server changes, and boards that were only ever on the member's own computer stay. If the member is added back, the next pull brings everything back. A backend that is unreachable or erroring hides nothing.

There are two independent causes behind a board-less member: the local pull above, and workspace access, which is granted per workspace rather than by account membership. A join that succeeds and still shows no boards may be the access side; the summary line says which.

Try again retries in place. When the invite screen cannot verify a link (a network error), its Try again button runs the invitation check again and shows the invitation once it answers. When the accept itself failed, Try again sends the accept again instead. Neither reloads the window. Four route-level specs ship NOT RUN on a machine without Postgres. guest-board-cap.test.ts, guest-route-cap.test.ts, guest-review4-cap.test.ts and guest-class-sweep.test.ts, in the hosted backend’s board test suite, drive the real routes against a real database. They cover the guest write refusal and comment success, a guest who owns a board being refused every sharing action and handed no join token, nav, My Work and search hiding a board the guest was never added to, the template, folder, dashboard and workspace paths, and a sweep that asserts a guest is refused on every write and admin route. GitHub CI runs Postgres; a developer machine may not. Their presence is not a pass.

First-run front door

What a person sees the first time they open Mission Control, before any board exists. It is a short series of screens rather than a form, and every one of them can be left for later.

The data question. The first screen asks the only thing that changes what happens next: "Got data somewhere else?" Answering it well (Asana, Monday, Jira, a spreadsheet) starts an import. Answering "Later, I just want to look around" skips the import for now and goes straight into the app — the question is asked once, not repeated.

The setup checklist. A short list of what Mission Control can do, answered with checkboxes. It is what the Guided setup reads to pick a starting point, and it replaces the older wizard's feature-checklist step.

The sample project. Choosing to look around first creates one small sample project, so the board, the left nav and Home all have something real in them instead of an empty state.

The comeback card. Home carries a card that offers to finish setup later. Answering it either resumes setup or dismisses it.

The short Guided setup. Four steps — AI comfort, the kind of software work, a project name, then one confirmation screen. It recommends a preset (Tasks / Projects / Operations / Engineering-Linear / Engineering-Jira) from those answers. The AI comfort answer can be changed any time afterward in Settings > Mission Control > AI comfort level; at Low (and when it was never picked) Mission Control sends nothing to an AI and its AI answers say so.

Leaving part-way is safe. Closing setup before the end marks nothing as done; the next visit offers "Pick up where you left off" and reopens at the step that was left. Setup is marked complete only once the starter project actually exists.

Coming from Tasks instead. A Tasks user pressing Try Mission Control enters at the setup checklist with boxes pre-ticked from their Tasks use, and gets a Your tasks step that copies lists into projects and/or turns on the Tasks side panel — see Moving from Tasks to Mission Control.

Role-based entry points (in development)

An extension to the setup wizard with three onboarding lanes and a 12-combination default surface matrix that picks a preset from two answers. It is gated behind a release feature that is off and internal-only, so customers never see a switch for it; when the gate is off the existing wizard behaviour is preserved.

  • Quick start — no questions. A short "Setting up your project" screen makes the starter project from what the checklist already answered, then opens it. Only when no preset was saved does the matrix pick one.
  • Guided setup — the short Guided setup described above. Closing it part-way, or continuing without a project after a failed save, keeps the answers so far and reopens on the first unfinished step.
  • Build with Omni — a 10–15 minute AI conversation that starts from the boxes ticked and the text typed on the checklist, and infers comfort, role and preset. Setup counts as finished only once that answer arrives; until then "Pick up where you left off" stays available. While the ticks lead here, the checklist notes that it runs on your API key and chosen model.

Which lane the checklist picks is a weighted rule, not a tick count: each box carries a weight, and a combination of many boxes, a high total weight, or a longer description picks the richer lane. The 12-combination matrix maps every comfort-and-role pair (3 comfort levels × 4 org roles) to a preset, a default view, an AI-behaviour level and a set of automatic features; empty answers default to medium comfort and a tasks role. Every Mission Control session is told the user's comfort, role and AI behaviour so it can adapt.

Admin-assigned onboarding (same machine): a manager can pre-fill the Guided setup answers for another user through the CLI (POST /pm/onboarding/assign). The target's Mission Control opens with those answers applied, and it is silently ignored if they already completed setup.

Peer inference is not implemented — there is no peer-inference feature in the tree today; the intended shape was to suggest features from team usage patterns, additive only.

Retake flow: "Redo Setup" in Mission Control's settings reopens the lane selector, not the wizard directly, when the gate is on.

Comment visibility (in development)

Item comments carry a visibility field, internal or external, defaulting to internal. Internal comments are visible only to board owners and members; external comments are visible to all roles including viewers and guests. The filter is applied at the query level, so viewers, guests and unauthenticated users see only external comments. A comment is a reply when it is created with a parent id, and the list route returns that parent id beside the raw row fields under the same name. Syncing from an external source preserves the existing visibility value, so a sync never overwrites a local choice.

The renderer component holding the visibility toggle and comment composer is built and unit-tested but not mounted anywhere, so no screen shows it yet; it was kept on purpose as a finished feature waiting for a place in the UI. The backend routes and role-based filtering are live. Board owners and members can change visibility on existing comments through the CLI. All three IPC channels are desktop-only.

Summarizing a comment thread

Summarizing a comment thread is a paid AI call — a small model over the thread's text, on demand, from the comment pane or from the CLI. It is metered, and the meter is a hard $0.50/day ceiling charged against its own cost label over a local-midnight day; there is no setting for it, so it does not appear in the settings reference or in Cost Control. If you summarize a lot in one day, the summaries simply stop being generated until the next day — nothing else in the comments feature is affected. This ships, unlike the comment-visibility gate above.

Item history, undo and rollback (unreleased)

Every Mission Control field change is tracked: item, column, old and new value, who did it (a human, an agent, an automation, an import or the API), the session and turn, and what caused it. Recording is fail-safe — it never breaks the change it is recording. An import records only actual changes. When a CLI edit is accepted in the cloud, the local copy records the rename, group move or archive as the same entries the offline fallback writes, so the later sync sees no difference and adds nothing. When a pull brings an item onto this computer for the first time, it records only that the item arrived and none of its cells, so a first download never reads as a burst of changes; the "what changed" answer dates that arrival by when the item was really created, not by when it was downloaded.

Undo reverts the last change for a field: a column value, a name, a create or a delete. Synced boards route the change through the Mission Control API and then update the local copy through authorised writers; local boards use those writers directly. Deletion undo restores from a snapshot. Every undo creates a new forward history entry linking back to the one it undid.

Rollback builds a plan showing which fields to revert plus conflict flags (a field changed more than once after the target point is flagged), then applies it with stale-plan detection — it refuses if the item was modified after the plan was made. Conflicts can be resolved by applying the revert or by keeping the current value for the conflicting fields only.

History is forward-only: undo and rollback add entries and never erase old ones. History is kept for 180 days. The item card has a History tab showing a timeline of changes with colour-coded actor badges, old-to-new value chips, relative timestamps and a per-entry Undo button.

A history entry that fails to save is no longer silent. Recording never breaks the change it is recording, so a failed write (usually a moment when the database is busy) used to leave only a log line, and the History tab looked complete while "who changed this" and Undo did not cover that edit. Each dropped entry is now counted, and one inbox card, "A board change is missing from its history", tells you. The next history write that saves clears the card, and the first save after a restart looks once for a card an earlier run left up, so the card never outlives the problem. Your boards and their data are unaffected.

Cross-entity linking

Board items can be linked to 14 entity types: Vault notes, mind maps, flowcharts, whiteboards, sessions, writer documents, URLs, email threads, meetings, GitHub pull requests, GitHub commits, GitHub branches and GitHub issues. Links are bidirectional.

  • The item card's Links tab groups linked entities by type with icons and a hover-reveal unlink action. GitHub commit links show the short SHA and a commit-message excerpt.
  • "Link to Board Item" in each visual tool's toolbar or context menu opens a search dialog over all boards and items.
  • Backlink chips appear in Vault, mind-map, flowchart and whiteboard headers and refresh when a link changes.
  • Create GitHub Branch on the Links tab creates a branch named after the item (slugified, e.g. feat/my-item-name) on the first GitHub-connected project, then links it back. It is gated behind an unreleased feature.
  • Clicking a GitHub commit link opens the commit on GitHub in the browser.

The GitHub commit scanner (in development) polls every five minutes for MC-N references in recent commit messages and creates commit links on the matching board items. It discovers repos from your project folder paths, fetches the last 100 commits per repo, and resolves references against the item number. Commit metadata is stored beside the link, and duplicates are prevented by unique constraints.

GitHub review requests (unreleased) add a background poller: when an item links a pull request, it fetches review state every ten minutes and shows a per-reviewer badge (approved, changes requested, pending, dismissed, commented). A state change raises an inbox alert with an Open item button that opens the linked item. A kill switch turns the poller off.

Item card Calendar and Gantt tabs

An item card can show that one item's plan as a calendar and as a Gantt chart, without switching the whole project to a Calendar or Gantt view.

  • What they show: the open item, its own sub-items, and the items on the same project linked to it through the Dependency column — what it waits on and what waits on it. Nothing else from the project appears. Rows run in a fixed order: what it waits on, the item, its sub-items, then what waits on it.
  • When they appear: the Gantt tab only on a project with a Timeline (or Week) column; the Calendar tab on a project with a Date or Timeline column. A project with neither keeps the usual tabs (Updates, Files, Activity Log, Links, Time, and History when that preview is on). They sit after the other tabs, and the arrow keys reach them like any other tab.
  • Same dates as the project: the Gantt tab reads the column the project's own Gantt view is set to (else the first Timeline column), and the Calendar tab uses the same rule as the project Calendar — the Timeline span when set, else the Date. Both calendars build their entries through one shared helper, so the tab and the project view never disagree.
  • Look-only: clicking a bar or a calendar entry opens that item's card; nothing can be dragged, edited or added from these tabs, and the Calendar tab carries no subscribe button. The Gantt keeps its zoom and critical-path controls.
  • Phone: the card is the same on the phone; on a narrow screen the Gantt opens with its left list folded so the chart has room.
  • Opened from My Work: the card loads the project itself (a recent saved copy when there is one); until it has the project the two tabs are simply absent.
  • Not here: Google Calendar, holidays and time off never show in these tabs. A plus button that lets each project pick which extra tabs its cards show is logged for later.

Goals

Hierarchical goal tracking scoped to a workspace. A goal has a title, description, status (on track, at risk, off track, achieved), an optional target date, a progress percentage, a progress mode (automatic or manual) and an optional parent goal for nesting. Goals can link to boards, and every change records an activity entry.

UI: a goals list page with status-filter pills and an empty state; a goal detail page with the status pill, progress bar, linked boards with a manage-boards modal, child goals and a paginated activity feed; a shared create-and-edit modal. Home surfaces active (non-achieved) goals in a collapsible summary widget showing up to five with inline progress bars and a "See all" link. The left navigation adds a Goals section between Dashboards and Trash. Parent-goal changes are checked for cycles.

Data: goals, goal-to-board links and goal activity live in the local database. The goal-comment routes are part of the same feature.

Portfolios

A named collection of boards scoped to a workspace. A portfolio has a title, a description and links to boards — simpler than goals, with no hierarchy, status or activity log.

UI: a portfolios list page with an empty state and a portfolio detail page with linked board rows, an unlink action and an inline board-linking picker. The left navigation adds a Portfolios section after Goals, whose See all button opens the list page.

Data: portfolios and their board links live in the local database. Board item-level data — status counts and health — is fetched from the backend rather than the local session bridge and is deferred to a follow-up.

Publish to Web

Boards and dashboards can be published as read-only web snapshots through the Omniscio Shares pipeline. The plugin renders a self-contained HTML document (inline CSS, no external resources, light or dark theme) and publishes it to cloud storage, producing a public URL.

What can be published: any of the board view types (table, kanban, calendar, chart, timeline, roadmap, form, workload, map) plus dashboards. A widget grid shows numbers, charts, batteries and tables; sprint-history widgets render placeholders, and form views show an explanatory empty state. Re-publishing unchanged content returns the same URL, and revoking follows the durable-intent pattern.

UI: a publish button in the board toolbar and in the dashboard header. Both open a modal with publish, update-snapshot and revoke-link controls. Password and expiry are set after publishing from the Shares panel, matching the Decks pattern. All three publishing actions are desktop-only.

When first-run setup appears

  • It waits until you are signed in. Signed out, Mission Control shows its sign-in screen and no setup screen; the "Got data somewhere else?" question opens by itself right after you sign in.
  • Invited teammates skip it. While an invite is open in Mission Control, setup stays closed. Accepting the invite marks setup finished, so an invitee never sees the data question, the checklist or a sample project. Declining leaves setup for the next time Mission Control opens.
  • Picking a tool is not finishing. Choosing a tool or "I have a file" on the data question opens the import; setup counts as finished only once an import actually creates at least one project. Leave the import part-way and the Home card offers to pick up at the data question.
  • Obsidian notes go to the Vault, not to a project. "I have notes in Obsidian" (desktop app only) opens the Vault's one-press Obsidian import on top of the data question, asking first to switch the Vault on if it is off. It does not finish setup or create a project; when the import closes, the data question is still open. See KMS Obsidian Import.
  • Settings → Mission Control → Finish setup switches to Mission Control's Projects tab with setup open where you left off.
  • The short spotlight tour waits. Its three tips ("Here's where your projects live", settings, Omni) appear only after setup is finished and no setup screen is open.

Setup wizard

The Guided setup wizard used to be a five-step interview; it is now the short Guided setup described under the first-run front door above. It is on by default and internal-only: customers have no switch for it, while dev and internal builds keep the Settings → Lab toggle as the team's off switch.

How to use it: enable Mission Control; the wizard is already on. With the front-door screens off, the wizard auto-opens when setup is incomplete and was never started. Each step saves on Continue. Closing it part-way saves nothing as done; it is remembered for "Pick up where you left off".

The final step shows the button is working while it runs. Setup is persisted as soon as the starter board exists, so the workspace defaults are a best-effort follow-up that can never leave setup stuck: a failure there still keeps the board, finishes setup and lands on it.

Preset recommendation: an engineering role gets Engineering-Linear by default (Engineering-Jira when the source tool was Jira); any early checklist item ticked gets Operations; the AI item is a confidence signal only and never triggers Operations; a team or department role with nothing ticked gets Projects; otherwise Tasks. The recommendation highlights one card but the user can override.

Presets: three core presets — Tasks (personal: 4 columns, table view), Projects (team: 7 columns, forms and team infrastructure) and Operations (department: 10 columns, all 8 views, dashboards, automations and formulas). Two engineering presets — Engineering (Linear) and Engineering (Jira) — add cycles, triage, sprints, epics and timeline. Eight platform presets (First-Agent-Board, Trello, Jira, Asana, Linear, Notion, ClickUp, Airtable) are assigned by the import flow, with single-platform imports mapping to the matching preset and multi-platform to Operations. All 13 presets control surface area, not access.

Mission Control's own settings page (Mission Control → your account menu → Settings) is one short page: How you work (your current preset with Change, and Customize — neither of which marks setup unfinished), Updates (what reaches your inbox, and the two automation alerts), Look (board density: Dense, Compact, Comfortable, Spacious; the theme always follows Omniscio's light or dark mode) and Document Import (import a CSV, JSON or Markdown file; column headers are always detected).

Omniscio Settings → Mission Control keeps only the switches that decide whether Mission Control is on and whether it reaches your inbox. A top card that is always open holds Enable Mission Control with an Open Mission Control button and a link to Mission Control's own settings; everything else sits in fold-up cards that start closed, each with a one-line summary. The cards are grouped into a Mission Control area (Sidebar, Notifications and quiet hours, Calendar and planning, Webhooks) and an Omni area (see ai-project-manager.md). Expand all opens every card, and the area chips jump to each group. The calendar due-date controls are in the Calendar panel's gear. Searching Omniscio Settings for a moved setting (preset, density, what goes to my inbox) opens Mission Control's page directly.

The setup checklist turns features on

The setup checklist ("What do you want Mission Control to manage for you?") is the one place you say what Mission Control should offer. It has 27 boxes. Ticking a box turns features on — a view, a menu, a column type in "+ Add column" — and a gray line under each box says what. It never adds a column, changes an item or deletes anything on a board.

  • Laid out as tabs. The categories (My Role, Basic Setup, Collaboration, Project Tracking, Power Features, AI & Agents) are listed on the left with how many boxes are ticked in each. Click one, or use the arrow keys, to see its boxes on the right, one per row. The first-time checklist and Customize look the same.
  • Customize (Mission Control settings → How you work) opens the same checklist with your current choices ticked. Apply to: All boards changes your default, which every board without its own choices follows. The line beside it says which setup style your ticks match.
  • One board. Pick a board in Apply to to give it its own choices. Only the boxes that turn something on are shown; boxes that differ from your default say only this board; Reset to my default removes the board's own choices. In the picker, a board with its own choices shows how many boxes differ.
  • What used to be Lab switches are now features the checklist turns on: sprints, approvals, automations (unticking it also stops that board's rules running, scheduled ones included), documents, spreadsheets and forms, the internal/shared choice on comments, email to tasks, Team Chat links and GitHub links (pull request, commit and branch links and review requests; the GitHub scanners only run while some board has this ticked).
  • Always on, no box: item history and undo, private saved views, and import.
  • The Tasks side panel has no checklist box either, but it is not always on. It has its own switch, Tasks side panel in Mission Control (Settings > Mission Control, off by default), and it also needs Tasks itself turned on.
  • When you first set up, the boxes you tick choose your default; if you skip, your setup style's boxes are used. When you update, each new box starts ticked exactly when its old Lab switch was on, and a board that showed sprints still does — nothing you had turned on disappears.
  • The lane counts are shares of the checklist. Ticking about two thirds of the boxes (18 of 27) or about a third with a long note (9 of 27) still picks the same setup path as 13 and 6 did on the old 20-box list.
  • CLI: the checklist can be read and set for your default or for one board, and resetting a board passes a null answer.
  • Mission Control — the overview page this continues.
  • Goals and Portfolios — the goals and portfolio surfaces in more detail.
  • Agent trust — the trust levels that decide what an agent may change on a board.

Last verified 2026-11-03