Cockpit & Pilotage
The cockpit is Stellary's strategic command center. It gives you a high-level view of your workspace's health — priorities, missions, decisions, and AI activity — so you can steer execution without getting lost in individual tasks.
What is the Cockpit?
The cockpit complements the kanban board by adding a strategic layer. Where the board manages execution (which tasks, in which column), the cockpit manages direction (which objectives, which priorities, which decisions).
Board
Execution: what, when, who
Cockpit
Strategy: why, where to, which priorities
Access the cockpit from the sidebar of your workspace. Access is controlled by a pilotage role (viewer, reviewer, or owner) that determines whether you can view, review AI proposals, or manage the full cockpit state.
Context Profiles
The cockpit adapts to what you need right now through three configuration dimensions. These control which sections are visible, how much detail is shown, and what data the AI focuses on when querying the cockpit context.
| Dimension | Options | What it controls |
|---|---|---|
| Profile | Pilotage, Execution, Revue | Your angle of view. Pilotage shows the full strategic overview. Execution focuses on todos and active missions. Revue highlights decisions and completed items. |
| Intent | Weekly Review, Daily Execution, Sprint Planning, Context Only | The purpose of your session. Determines which data is prioritized and how the AI frames its suggestions. |
| Detail Level | Minimal, Standard, Detailed | How much information to display. Minimal shows key metrics only. Standard shows full lists with basic details. Detailed includes traces and history. |
Your preferences are persisted in localStorage. When an AI agent queries the cockpit, it selects the appropriate profile for its task automatically — you don't need to configure anything for agent access.
Priorities
Priorities are your team's current focus areas — the 3 to 4 things that matter most right now. This strict work-in-progress limit keeps the team aligned and prevents context-switching.
- Create a priority with a module (scope) and a label (focus area)
- Maximum of 3–4 active priorities at any time to enforce focus
- Status lifecycle:
active→completedordeferred - Displayed in a grid within the cockpit, visible to the entire team
Missions
Missions are strategic goals linked to your projects. They represent the higher-level objectives that your day-to-day tasks contribute to.
- Create missions with an objective, a module, and a linked project
- Status lifecycle:
active→completedorblocked - When blocked, provide a reason to give the team visibility on what's holding things up
Dependency
Waiting on another team or system
Validation
Needs approval before proceeding
Resource
Lacking people or budget
Decision
Waiting on a strategic choice
Todos
Todos are concrete action items linked to missions and/or priorities. They bridge the gap between strategic objectives and day-to-day execution.
- Quick-create from the cockpit with a title and optional description
- Link to a mission or priority to maintain traceability
- Status lifecycle:
active→completedorblocked - Track completion progress with a done/total counter for grouped tasks
Decisions
Decisions capture the strategic trade-offs your team makes. Recording them with their context and impact ensures the reasoning behind choices is preserved and searchable.
- Record the title, status, and impact assessment
- Track metadata: decided by (who made the call) and decided at (when)
- Searchable history accessible to the entire team — no more "why did we do this?"
Active Documents
Active documents are workspace documents that are actively referenced in the current planning cycle. Unlike a wiki or document library, this section only surfaces what matters right now.
- Link workspace documents to the cockpit context
- Add a title, an owner, and a source type
- Keep the list lean — only what the team needs to reference this cycle
AI Proposed Actions
When an AI agent proposes an action (create a card, move a task, flag a blocker), it enters a review queue visible in the cockpit. Nothing is applied without explicit human approval.
Each proposed action follows a clear workflow:
Suggested
AI proposes an action
Approved
Reviewer validates the proposal
Applied
Action is executed
Rejected
With a reason (presets available)
Approve
Queue the action for execution
Approve & Apply
Validate and execute in one click
Reject
With a predefined or custom reason
This is Stellary's core principle: AI proposes, humans decide. No action is executed without explicit validation from a reviewer. Rejection reasons are tracked so the AI learns from feedback.
AI Agents & Missions
AI agents can run missions on cards — autonomous or supervised tasks that stream progress in real-time. The cockpit gives you visibility into active agent work across the workspace.
Mission Lifecycle
Queued
Mission created, waiting for agent
Running
Agent is actively working
Awaiting Approval
Supervised agent needs human review
Completed
Mission finished successfully
Failed
Something went wrong
Cancelled
Stopped by a user
- Missions run on cards — assign an AI task from any card in your board
- Progress streams in real-time via SSE, so you see each step as it happens
- Supervised agents propose changes before execution — you approve or reject each proposal
- Cancel any running mission at any time if the agent goes off track
Automations
Automations are project-level rules that execute actions when specific events occur. Define a trigger, set optional conditions, and let Stellary handle the rest.
Available Triggers
| Trigger | Fires when |
|---|---|
task_moved | A card is moved to a different column |
task_created | A new card is created |
checklist_complete | All checklist items on a card are checked |
due_date_reached | A card's due date arrives |
task_status_changed | A card's status changes (open, in progress, done, cancelled) |
task_assigned | A card is assigned to a member |
task_added_to_board | A card is added to a scope or board |
- Create rules from the Automations panel in the board view (lightning bolt icon)
- Each rule has a trigger, optional conditions (JSON filter), and an action to execute
- Rules can be enabled or disabled individually without deleting them
Automation Traces
Every automation execution and AI action is recorded in an audit timeline. This gives you a complete picture of what happened, who triggered it, and what was affected.
- Each trace records the actor (human or AI), the target (card, mission, etc.), and a precise timestamp
- Results include success or failure status for every execution
- View the last 50 traces directly in the cockpit to audit recent activity
- Traces are available in the Detailed context profile level for full visibility