Documentation Index
Fetch the complete documentation index at: https://docs.archal.ai/llms.txt
Use this file to discover all available pages before exploring further.
TL;DR
- Add
./.archal/harness.ts— readsARCHAL_ENGINE_TASK, calls your agent, prints the result. - Add
.archal.jsonwithagentpointing at it. archal run --task "..." --twin github.
1. Add a headless harness
Archal needs a runnable entrypoint it can spawn without booting the full app shell. The harness should:- short-circuit when
ARCHAL_PREFLIGHT=1soarchal runcan validate the entrypoint before provisioning twins: - read
ARCHAL_ENGINE_TASK - call your agent runtime
- print the final result to stdout as a JSON object like
{"text": "..."}and keep logs on stderr. Archal extracts structured output in preference to plain text — see Your first harness for the full stdout contract.
2. Check the harness
archal run preflights automatically. To reproduce a preflight failure outside Archal:
3. Add .archal.json
Optional with --harness, but useful as a project default:
4. Service traffic
Your harness should use normal SDKs and service domains. Archal’s proxy routes supported service traffic to the scenario twins without exposing twin URLs to the agent process.5. Skip Docker unless you need it
Repo-local harnesses run locally by default. Pass--docker only when the runtime depends on container-only system libraries.
6. Run the first task
7. Promote to a scenario
8. Find the artifacts
.archal/cache/last-run.json and .archal/cache/runs/*.json. Use --output json only for machine-readable stdout.
