corral-agent

Generated by scripts/gen-cli-docs.sh from corral-agent’s own -h output and its main.go doc comment. Do not hand-edit — run scripts/gen-cli-docs.sh and commit the result.

Usage

corral-agent — reference LLM-driven audit-role worker (no default backend or model)

Usage:
  corral-agent            connect to the brain and work the queue
  corral-agent --version  print the build version and exit
  corral-agent -h         print this help and exit

Env:
  CORRAL_BRAIN       brain URL (default http://127.0.0.1:9019/mcp/)
  AGENT_ROLE         role(s) to serve (default generalist): a single role
                     (mutant-generator | test-writer | test-critic | ...), a
                     comma-separated list to claim any ready task in that set,
                     or "generalist"/"any"/"*"/empty to claim ANY ready task
  AGENT_NAME         display name in the swarm UI (default same as AGENT_ROLE,
                     e.g. "researcher+designer" or "generalist")
  AGENT_WORKSPACE    working directory for edits (default $TMPDIR/corral-demo-ws)
  MODEL_BACKEND      REQUIRED: ollama | openai (any OpenAI-compatible endpoint) | openrouter | gemini | anthropic (alias claude)
  AGENT_MODEL        REQUIRED: model name passed to the backend
  OLLAMA_URL         REQUIRED with MODEL_BACKEND=ollama: the daemon's URL (no default address)
  CLOBBER            set "1" to ignore coordination conflicts and edit anyway (demo of what NOT coordinating looks like)