API Channels
CLI chat, OpenAI-compatible HTTP, and A2A integration paths.
Not every integration needs a human chat platform. Facio also exposes API-style entry points for scripts, internal tools, and other agents.
CLI chat
The Facio CLI can send a direct message to the agent loop with an explicit session ID. It is useful for local operator testing and debugging, but the Docker quickstart is the primary deployment path.
Useful options include:
| Option | Meaning |
|---|---|
--message / -m | Message to send. |
--session / -s | Session key, default cli:direct. |
--workspace / -w | Workspace directory. |
--config / -c | Config file path. |
--markdown / --no-markdown | Render assistant output as Markdown. |
--logs / --no-logs | Show runtime logs during chat. |
OpenAI-compatible HTTP
The /v1/chat/completions path lets OpenAI-style clients call Facio. See API Gateway for request shape and limits.
Use it for automation that needs a simple request/response agent endpoint. Use Placet when the workflow needs rich approvals, files, forms, or management controls.
A2A
A2A is opt-in and exposes an authenticated inter-agent RPC surface plus a public agent card.
| Setting | Meaning |
|---|---|
FACIO_A2A_ENABLED | Mount /a2a/v1/* and /.well-known/agent-card.json. |
FACIO_A2A_TOKEN | Bearer token for A2A peers. |
| Management token | Also grants access to A2A when mounted. |
Management API endpoints under /api/v1/a2a/peers can store peer aliases, fetch cards, delete peers, and call a peer.