Documentation Index
Fetch the complete documentation index at: https://ade-app.dev/docs/llms.txt
Use this file to discover all available pages before exploring further.
What a Mission is
A Mission is ADE’s orchestration layer for work that outgrows a single chat thread. Where Agent Chat is a fast loop in one lane, a Mission turns a bigger goal into phases, delegates execution to workers running in their own lanes, pauses at checkpoints when you or the plan require it, and gathers every log, summary, and artifact into one reviewable record.Missions are visible as a tab in every build so teams can plan around them, but creation and live runs are paused in production builds until the orchestration flow is production-ready. Dev and internal builds run Missions end-to-end today. The tab shows a “Coming soon” preview in packaged app releases — plan with it now, automate with it when it ships.
Mission planning
Turn a larger goal into phases, worker lanes, checkpoints, and proof requirements before any agent starts typing.
Lane orchestration
Coordinate multi-agent work across isolated lanes, then bring results back with reviewable handoffs.
Human checkpoints
Pause when ADE needs operator input, budget approval, or a decision before external side effects happen.
Artifacts and closeout
Collect logs, summaries, screenshots, and implementation proof into one mission record, attached to the associated lanes and PRs.
When a Mission is the right tool
| Use a Mission when… | Use Chat or an Automation when… |
|---|---|
| Work spans several lanes or files that should not step on each other | A single file or targeted change resolves the task |
| Human approval should gate a destructive or external step | No external side effects, or you will watch the whole run |
| You need a durable audit of plan, decisions, and proof | A transcript in Chat or History is enough |
| A plan should be reviewed and edited before agents spend tokens | The prompt fully describes the work |
Phases
Every Mission moves through an ordered sequence. The boundary between phases is a checkpoint — ADE snapshots state before advancing so the mission can be paused, resumed, or replayed.Planning
The orchestrator reads the goal, picks a strategy, lays out worker assignments and lane routing, and estimates cost and tool needs. You can open the Planning Prompt Preview to inspect — and optionally edit — the exact prompt the planner receives.
Approval gate
The finished plan is presented before any worker spends tokens. This is a hard stop. Approve, request revisions with feedback, or cancel while nothing has run.
Execution
Workers run according to the approved plan. The Plan tab surfaces worker status, token consumption, and the live decision feed. You can talk to the orchestrator in the Chat tab at any time without interrupting execution.
Worker strategies
Before planning starts, the planner picks how workers should run. The goal is to shorten wall-clock time without creating merge pain between workers.Sequential
Sequential
One worker at a time. Output from step N is available as context for step N+1. Used when steps have tight dependencies.
Parallel, limited
Parallel, limited
A small, bounded number of independent workers run at once. The planner sets a cap to avoid rate limits and context overlap.
Parallel, fan-out
Parallel, fan-out
Every independent unit starts at the same time in its own lane. Fastest for work that genuinely factors into independent chunks (for example, adding tests for a batch of functions).
Where Missions live
- Missions tab — the canonical surface for creating, reviewing, and monitoring missions. Always visible in the sidebar. In packaged builds the tab shows a “Coming soon” preview with the feature summary.
- Automations — an automation rule can select
missionas its execution surface. In packaged builds the mission option is visible but disabled with a “Coming soon” chip, so teams can see how they would wire mission automations and plan around it. - Chat — any chat thread can escalate into a mission; conversation history seeds the planner.
Continue reading
Creating missions
Start a mission from the Missions tab, a lane, the CTO chat, or an existing chat session.
Workers and interventions
How workers run, when missions pause for human input, and how failures are classified.
Budget and permissions
Budget caps, per-provider permission modes, proof capture, persistence, and troubleshooting.