MCP and HTTP
MCP stdio
agent-exec mcp exposes run, status, tail, wait, and kill. A disconnected client does not stop the managed job; reconnect with its job_id.
AGENT_EXEC="$(command -v agent-exec)"
"$AGENT_EXEC" mcpHTTP server
agent-exec serve binds to 127.0.0.1:19263 by default. --port changes only the loopback port. A non-loopback --bind requires --insecure and a nonempty AGENT_EXEC_SERVE_TOKEN.
The token uses Authorization: Bearer authentication only for POST /exec and POST /kill/{id}. It does not protect read endpoints, so keep the default loopback bind or add a firewall, private network, or authenticating reverse proxy. Do not expose the server directly to the public internet.
--allow-origin permits one explicit CORS origin. The wildcard * is rejected.
| Method | Endpoint | Equivalent |
|---|---|---|
| GET | /health | Health envelope |
| POST | /exec | run |
| GET | /status/{id} | status |
| GET | /tail/{id} | tail |
| GET | /wait/{id} | wait --forever |
| POST | /kill/{id} | kill |
Documented handler responses use the CLI schema_version, ok, and type envelope. Router-level unmatched paths and methods are outside this response contract. POST /exec requires a nonempty command array of strings. Optional fields are cwd (string), env (string-to-string object), timeout (seconds), wait (default true), until (default 10), and max_bytes (default 65536). Unknown fields are rejected.
curl -X POST http://127.0.0.1:19263/exec \
-H 'Content-Type: application/json' \
-d '{"command":["sh","-c","echo hello"]}'