Start Agent Session
Start a new agent session. Provide at most one of config_id, config, or template_id; with none, the customer's default-config ladder resolves the config. 400 when idle_start is combined with prompt or handoff, or when a handoff is combined with any config source or override.
Start a new agent session.
Provide at most one of config_id, config, or template_id; with none, the customer's default-config ladder resolves the config. 400 when idle_start is combined with prompt or handoff, or when a handoff is combined with any config source or override.
Authorization
HTTPBearer In: header
Header Parameters
Request Body
application/json
TypeScript Definitions
Use the request body type in TypeScript.
Parameters for starting a new agent session.
Response Body
application/json
application/json
curl -X POST "https://example.com/v1/sessions" \ -H "Content-Type: application/json" \ -d '{}'{ "agent_config_id": "string", "attributed_user": { "avatar_url": "string", "id": 0, "login": "string", "type": "User" }, "attribution_id": "string", "attribution_type": "github_user", "client_version": "string", "config_override": {}, "context_repository": "string", "cost_fee": 0, "cost_sandbox_cpu": 0, "cost_sandbox_memory": 0, "cost_tokens": 0, "created_at": "2019-08-24T14:15:22Z", "customer_id": "string", "default_resolution": "repo_default", "exit_status": "completed", "harness": "claude_code", "id": "string", "last_activity_at": "2019-08-24T14:15:22Z", "last_message_at": "2019-08-24T14:15:22Z", "live_summary": "string", "metadata": {}, "output_prs": [ { "gh_pr_id": 0, "number": 0, "repo_full_name": "string", "title": "string", "url": "string" } ], "parent_agent_session_id": "string", "parent_kind": "continuation", "prompt": "string", "prompting": { "blocked_reason": "mention_surface", "detail": "string", "enabled": true, "surface_name": "string" }, "replayed_from_session_id": "string", "resolution_source": "repo_default", "resolved_budget_cents": 0, "resolved_budget_source": "system", "resolved_config_name": "string", "sandbox_id": "string", "session_key": "string", "session_state": "idle", "source": "react", "status": "scheduled", "status_reason": "string", "stopped_at": "2019-08-24T14:15:22Z", "stopped_by": 0, "stopped_by_user": { "avatar_url": "string", "id": 0, "login": "string", "type": "User" }, "surface": { "run": "scheduled", "session": "alive", "status": "string" }, "tokens_cache_creation": 0, "tokens_cache_read": 0, "tokens_input": 0, "tokens_model": "string", "tokens_output": 0, "tokens_total": 0, "updated_at": "2019-08-24T14:15:22Z", "webhook_id": "string", "webhook_type": "github.issue"}List Agent Sessions
List the customer's agent sessions, optionally filtered by config, source, time window, and attributed author.
Search Sessions
Search the customer's sessions, grouped by session, over everything a session left behind: step text, recap text, created PRs, and recap-embedding similarity. This is how an agent answers "did Tony look into X?": resolve the author via /github/members, search, then read the winning session via /sessions/{id} (recap) and /sessions/{id}/records (transcript).