Deploy via REST API

Authentication

The Free plan does not include API keys; they are on Pro and up.

Create an org API key (sk_live_...) with the Agent role in Settings → Security
Exchange it for a JWT via the auth API (it lasts 1 hour)
Call the agent API with Authorization: Bearer <jwt> and X-Agent-Id

That token can start a run of its own agent and follow or stop it, and nothing else. See Authentication → for details and the full list.

Run your agent

curl -X POST https://agent-api.auteryn.ai/api/run \
  -H "Authorization: Bearer eyJ..." \
  -H "X-Agent-Id: your-agent-id" \
  -H "Content-Type: application/json" \
  -d '{"messages":[{"role":"user","content":"Summarize the open P1 bugs"}],"stream":true}'

import httpx, json

with httpx.stream(
    "POST",
    "https://agent-api.auteryn.ai/api/run",
    headers={
        "Authorization": "Bearer eyJ...",
        "X-Agent-Id": "your-agent-id",
    },
    json={
        "messages": [{"role": "user", "content": "Summarize the open P1 bugs"}],
        "stream": True,
    },
    timeout=None,  # the stream can go quiet for longer than httpx's 5 s default
) as r:
    for line in r.iter_lines():
        if line.startswith("data: "):
            event = json.loads(line[6:])
            print(event)

// stream: false → 202 Accepted; follow up via the poll/stream endpoints.
const response = await fetch("https://agent-api.auteryn.ai/api/run", {
  method: "POST",
  headers: {
    Authorization: "Bearer eyJ...",
    "X-Agent-Id": "your-agent-id",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    messages: [{ role: "user", content: "Summarize the open P1 bugs" }],
    stream: false,
  }),
});
const { task_id, id: run_id } = await response.json(); // 202
// Then poll GET /api/runs/{run_id}/events or open GET /api/tasks/{task_id}/stream.

Streaming events

When stream: true, the response is Server-Sent Events using the AG-UI protocol. Each data: frame is one JSON event with a type (and snake_case fields). Core types include RUN_STARTED, TEXT_MESSAGE_START / TEXT_MESSAGE_CONTENT / TEXT_MESSAGE_END, TOOL_CALL_START / TOOL_CALL_ARGS / TOOL_CALL_END / TOOL_CALL_RESULT, RUN_FINISHED, and RUN_ERROR. Agent-specific signals (reasoning, artifacts, usage, and USER_INPUT_REQUIRED for human-in-the-loop) arrive as CUSTOM events. See the full AG-UI event-type reference →.

Continuing a conversation

Pass a task_id to maintain context across multiple requests (an agent token can continue only conversations it started):

{
  "messages": [{"role": "user", "content": "Now focus on only the critical ones"}],
  "task_id": "task_xyz789"
}