# Start a session

> Start a session: one run of your own harness (a Claude Code session, a Codex run) on a repository, so others can see who is working on what.

<div class="g1t-endpoint"><span class="g1t-method" data-method="post">POST</span><code>/repos/{owner}/{name}/sessions</code></div>

`harness` is claude_code, codex, cursor, copilot or other. Give `harness_session_id`, the harness's own id for the session, and starting again with it returns the session already started (refreshing the issue, pull request, branch, title and head given) instead of a new one, so a hook that fires twice is harmless. `issue` and `pull` attach it to an issue and a pull request of the repository; a pull request for an issue attaches that issue too. A subagent is a session of its own: give `parent_id` (one of your sessions here), its `agent_type`, the `task` and `prompt` its parent gave it, and as `harness_session_id` the parent's harness id, `/agent/` and the subagent's id; it starts on the parent's issue, pull request and branch unless it says otherwise, sees and is seen by its siblings in [`get_work`](/reference/api/claims-and-work/get-work/) and [`list_sessions`](/reference/api/sessions/list-sessions/), and once ended stays ended when started again. Returns the session (`id` `ses_…`, `state` active, idle or ended). A new session, or one started again after it ended, is a session.started event. Needs a confirmed email address and a role that may open issues.

Idempotent on `harness_session_id`: a second start with it returns this session, so call it from a hook without checking first. `state` is `active`, `idle` (no heartbeat for 15 minutes) or `ended` (ended, or no heartbeat for 2 hours: `end_reason` `timed_out`). A subagent gives `parent_id`, `agent_type`, `task` and `prompt`. See [Sessions](/guides/sessions/).

- **Authentication:** Required. Send an [access token](/reference/api/#authentication) as `Authorization: Bearer`.
- **MCP tool:** [`session`](/reference/mcp/#session) with `action` `start`, and the same inputs
- **Scope:** An access token needs [`pull_requests:write`](/guides/authentication/#scopes).

## Path parameters

| Name | Type | Required | Description |
| --- | --- | --- | --- |
| `owner` | string | Yes | The workspace that owns the repository. |
| `name` | string | Yes | The repository's name. |

## Body parameters

Send a JSON object. Names are `snake_case`, as in responses; the `camelCase` spelling is accepted too.

| Name | Type | Required | Description |
| --- | --- | --- | --- |
| `issue` | integer | No | The issue it works on. |
| `pull` | integer | No | The pull request it works on. |
| `branch` | string | No | The branch it works on. |
| `title` | string | No | What it is doing, in a few words. Up to 200 characters. |
| `head_commit` | string | No | The commit it is at, such as after a push. |
| `harness` | string | Yes | The harness the session runs in. One of `claude_code`, `claude_managed_agents`, `codex`, `cursor`, `copilot`, `other`. |
| `harness_session_id` | string | No | The harness's own id for the session, such as Claude Code's session_id; for a subagent, the parent's, `/agent/` and the subagent's id. Starting again with it returns the same session. |
| `parent_id` | string | No | For a subagent: the session that started it (ses_…), one of yours on the repository. |
| `agent_type` | string | No | For a subagent: its kind, as the harness names it (Explore, security-reviewer). |
| `task` | string | No | For a subagent: the task it was given, in a few words. Up to 200 characters. |
| `prompt` | string | No | For a subagent: the prompt its parent gave it. Kept up to 4,000 characters. |
| `model` | string | No | The model it runs on, when known. |

## Example request

```sh
curl -X POST https://api.g1t.sh/repos/flagon-io/hello/sessions \
  -H "Authorization: Bearer $G1T_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "harness": "claude_code",
    "harness_session_id": "5b0c1f9e-3d2a-4e8b-9f61-0c7d2a9e4b13",
    "issue": 12,
    "branch": "greet-caller",
    "title": "Make the greeting name the caller"
  }'
```

## Example response

A successful request answers `200` with:

```json
{
  "id": "ses_01m4b2c3d4e5f6g7h8j9k0m1n2",
  "repo_id": "rep_01m3m5q6p0e2qaw6mmjahk0qrr",
  "issue": 12,
  "pull": null,
  "branch": "greet-caller",
  "actor": {
    "id": "usr_01kkntcg1eeb98j62xjm7eh09p",
    "username": "syntaqx"
  },
  "harness": "claude_code",
  "harness_session_id": "5b0c1f9e-3d2a-4e8b-9f61-0c7d2a9e4b13",
  "title": "Make the greeting name the caller",
  "state": "active",
  "head_commit": null,
  "started_at": "2026-10-12T14:02:11.204Z",
  "last_seen_at": "2026-10-12T14:02:11.204Z",
  "ended_at": null,
  "end_reason": null
}
```

## Errors

A failed request answers with one of these statuses and a body like `{"error": {"code": "not_found", "message": "Repository not found."}}`. See [errors](/reference/api/#errors).

| Status | Code | When |
| --- | --- | --- |
| 401 | `unauthenticated` | A token is required, or the one sent is not valid. |
| 403 | `forbidden` | The token is valid but not allowed to do this, such as a member-only change, or a token without the scope it needs. |
| 404 | `not_found` | It does not exist, or you cannot see it. |
| 409 | `conflict` | The request conflicts with the current state. |
| 422 | `invalid` | The input is not valid. `message` says which field and why. |