Today · Missions
Missions
A mission is one graph run across many agents, treated as a single accountable unit. RepoOps runs no graph. It reads the run events your multi-agent tools already produce and remembers the mission: the agent nodes, the edges between them, when each node wrote to shared state, where two nodes contended, a cost rollup over every node, and one outcome receipt answering whether the organization achieved the goal and what it cost.
Where to find it
- Local tab:
/missions.html?repo=<id> - List API:
GET /api/missions?repo=<id> - Detail API:
GET /api/missions/<id>?repo=<id> - Ingest API:
POST /api/missions/events?repo=<id> - On-disk log:
.claude/brain/missions/<repoId>.jsonl (append-only, under the aggregator's brain root)
What it does for you
See the whole agent graph, not one agentThe detail view draws the mission's nodes (agents) and the edges between them, each node badged running, done, or failed. The overall mission status is derived from its nodes: any failed node makes the mission failed, all done makes it done, otherwise it stays running.
Read the shared-state timeline and catch contentionEvery shared-state write is listed oldest-first with the node that wrote it. Any path written by more than one distinct node is flagged as contention, so you can see where two agents touched the same file.
Get a cost rollup with no fabricated numbersThe mission cost is the sum of the per-node spend the telemetry already captured. It adds no new meter: a node with no recorded cost reads honest-null, never $0, and the receipt reports how many nodes are unpriced.
Read one outcome receipt for the runEach mission carries a receipt: the verdict (verified only when the mission passed against a declared contract, otherwise completed), how many nodes ran out of the total, the cost, and whether it is signed. It reflects the folded outcome; RepoOps never re-executes the graph to judge it.