Synopsis
# Solo (one agent)
aigon feature-autonomous-start <feature-id> <agent> [--review-agent=<agent>] [--stop-after=implement|review|close]
# Fleet (multiple agents)
aigon feature-autonomous-start <feature-id> <agent1> <agent2> [--eval-agent=<agent>] [--stop-after=implement|eval|close]
# Status check
aigon feature-autonomous-start status <feature-id>Shortcuts
/aigon:feature-autonomous-start(slash command)
Description
Starts the AutoConductor — a detached tmux session that monitors the workflow engine and triggers the next lifecycle stage automatically. The mode it runs depends on the number of implementation agents:
- Solo (1 agent): implement → optional review → close
- Fleet (2+ agents): implement in parallel → evaluate → close
When finished, the AutoConductor kills its own tmux session so the dashboard badge clears.
Flags
| Flag | Mode | Description |
|---|---|---|
--review-agent=<agent> | Solo only | Agent to run the code review after implementation. If omitted, skips review. |
--eval-agent=<agent> | Fleet only | Agent to run evaluation. Defaults to the first implementation agent. |
--stop-after=<value> | Both | Where to stop — see table below. Default: close. |
--stop-after values
Solo mode
| Value | Behaviour |
|---|---|
implement | Stop after agent submits |
review | Stop when review session starts (review is still running; close manually) |
close | Default. Implement → review (if --review-agent set) → close |
Fleet mode
| Value | Behaviour |
|---|---|
implement | Stop after all agents submit |
eval | Stop after evaluation starts (pick winner and close manually) |
close | Default. Implement → eval → auto-close the winner |
Usage
# Solo — implement and close, no review
aigon feature-autonomous-start 42 cc
# Solo + review — implement, gg reviews, then close
aigon feature-autonomous-start 42 cc --review-agent=gg
# Solo + review — stop after review starts, close manually
aigon feature-autonomous-start 42 cc --review-agent=gg --stop-after=review
# Fleet — cc and gg implement, cc evaluates, auto-close winner
aigon feature-autonomous-start 42 cc gg --eval-agent=cc
# Fleet — stop after eval, pick winner manually
aigon feature-autonomous-start 42 cc gg --eval-agent=cc --stop-after=eval
# Implement only, then hand back control
aigon feature-autonomous-start 42 cc --stop-after=implement
# Check AutoConductor status
aigon feature-autonomous-start status 42From the dashboard
Click Start Autonomously on any backlog feature card. The modal adapts based on agent count: one agent shows the Reviewer picker; multiple agents show the Evaluator picker.
AutoConductor session
Runs as a detached tmux session: <repo>-f<id>-auto. Poll interval: 30 seconds.
# Attach to watch the log live
tmux attach -t brewboard-f42-auto
# Check status without attaching
aigon feature-autonomous-start status 42Solo + review poll log
[09:15:17] state=implementing agents=[cc=running] allReady=false ...
[09:16:47] state=implementing agents=[cc=ready] allReady=true ...
🚀 Spawning review session: brewboard-f42-review-gg-...
✅ Started review session: brewboard-f42-review-gg-...
[review-start 1/10] running=true
✅ Review started. Waiting for review agent to complete...
[review-close 1/120] reviewCompleted=false reviewState=in-progress
[review-close 4/120] reviewCompleted=true reviewState=idle
✅ Review completion signaled. Proceeding to close.
🚀 Triggering: aigon feature-close 42
✅ Feature closed. AutoConductor finished.Fleet + eval poll log
[09:15:17] state=implementing agents=[cc=running, gg=running] allReady=false ...
[09:17:47] state=implementing agents=[cc=ready, gg=ready] allReady=true ...
🚀 Eval session name: brewboard-f42-eval-cc-...
✅ Started eval session: brewboard-f42-eval-cc-...
[eval-start 1/10] state=evaluating running=true
✅ Evaluation started. Waiting for eval agent to complete...
[eval-close 1/120] state=evaluating winner=none closeTriggered=false
[eval-close 3/120] state=evaluating winner=cc closeTriggered=false
🏆 Winner: cc — closing feature 42
✅ Feature closed autonomously. AutoConductor finished.Review completion signal
In solo+review mode, the AutoConductor waits for an explicit review-complete signal, not just for the session to exit. The review agent emits this automatically via the shell trap when it exits cleanly (exit code 0).
If the review session exits without signaling, the AutoConductor errors and prints manual recovery instructions.
Resilience
If the AutoConductor dies mid-run, the feature continues in whatever stage it reached. Complete manually:
aigon feature-review 42 # if review hasn't run (solo)
aigon feature-eval 42 # if agents haven't been evaluated (fleet)
aigon feature-close 42 # merge (solo)
aigon feature-close 42 cc # merge the winner (fleet)Related
feature-start— manual Fleet/Drive setupfeature-review— manual code reviewfeature-eval— manual evaluationfeature-close— manual close- Autonomous Mode guide