Creating a Mission
**Screenshot: “New Mission” dialog — show the full creation form with fields for title, description, lane selector, model picker, tool permissions grid, and budget cap input.
From the Missions tab
From a Lane
From the CTO chat
From an existing chat
Open Missions
Click Missions in the left sidebar. You’ll see your mission history list.
Click New Mission
Press the + New Mission button in the top-right corner of the Missions page.
Fill the creation form
Complete the mission creation dialog:
- Title — short, descriptive name (shown in the mission list and history)
- Description — detailed prompt describing the goal; this is what the orchestrator receives
- Lane(s) — one or more lanes to scope the mission to; workers run in these lanes’ worktrees
- AI Model — the model to use for the orchestrator and workers
- Tool permissions — which ADE tools workers can invoke; defaults to the lane’s tool palette
- Budget cap — optional token or USD ceiling; mission pauses if the limit is approached
Review preflight checks
Before the mission starts, ADE runs preflight checks: verifies the selected lanes are available, confirms the configured model is reachable, validates the permission config, and warns if any required computer-use capabilities are unavailable.
Launch
Click Start Mission. The mission enters the Planning phase immediately.
Open a Lane, then click Start Mission in the lane toolbar. The creation dialog opens pre-filled with that lane’s context and tool palette.
In the CTO chat panel, describe your goal and say “start a mission for this.” The CTO agent creates and configures the mission, pre-filling the description from the conversation context. You’ll see the preflight dialog before planning begins.
In any Agent Chat session, say “start a mission for this task.” ADE transitions the current context into a new Mission, carrying over the conversation history as seeding context for the planner.
The Mission Detail View
Once a mission is running (or completed), click it in the mission list to open the detail view. The detail view has four tabs:
**Screenshot: Mission detail view with the Plan tab open — show worker cards with status dots, token counters, the execution strategy badge, and the live activity feed at the bottom.
Plan tab
Chat tab
Artifacts tab
History tab
The Plan tab is the mission’s live control room. It shows:
- Execution strategy badge — which strategy the meta-reasoner selected
- Worker cards — one card per spawned worker, each with: assigned lane, current status (running / completed / failed / blocked), elapsed time, and running token count
- Phase status bar — visual progress through Planning → Execution → Validation
- Live activity feed — a chronological stream of meaningful events: worker starts, key decisions, tool calls, phase transitions, and intervention triggers
- Open interventions — any pending approval requests or questions that need your response
Workers that have intervention_required status will highlight in amber. Click the intervention card to respond. Chat directly with the mission’s orchestrator while it runs. Use this to:
- Ask the orchestrator why it made a planning decision
- Provide additional context mid-mission (“the API endpoint changed to v2”)
- Redirect a worker that is going off-track
- Request a status summary without digging through the activity feed
Chat signals from worker streams are filtered by ADE’s noise detector, which suppresses repetitive status updates. Only meaningful signals — decisions, blockers, completions — surface to the mission chat. All outputs produced during the mission are collected here:| Artifact type | Description |
|---|
patch | Code changes (diffs) committed by workers |
pr | Pull requests created during execution |
test_report | Test run output from the validation phase |
screenshot | UI screenshots taken by workers via computer use |
browser_verification | Browser-based proof artifacts from computer-use steps |
browser_trace | Full browser action traces |
plan | The generated execution plan (snapshot at approval gate) |
summary | Orchestrator-generated mission summary |
note | Inline notes attached by the orchestrator or by you |
A full, immutable timeline of everything that happened inside the mission — every planning step, tool call, model decision, worker message, and phase boundary. The History tab is the source of truth for auditing and debugging.Events are timestamped and categorized. You can filter by worker, by event type, or by phase.
