Hatchet

Hatchet provides durable workflow execution for background tasks, offering fan-out/fan-in orchestration, retry logic, step dependencies, and worker management via an SDK-first approach. For agents, the core integration pattern is: define workflows as code (TypeScript/Python/Go), register workers that execute steps, and trigger workflow runs via the API or SDK. The REST API exposes workflow run triggers, status polling, and event ingestion — useful for agents that need to queue durable background work and track completion. Simpler operational model than Temporal (fewer components, easier self-hosting). Cloud offering available at cloud.onhatchet.run. Confidence is docs-derived.

REST API at <instance>/api/v1. Key endpoints: POST /api/v1/workflows/{id}/trigger (trigger a run), GET /api/v1/workflow-runs/{id} (poll status), POST /api/v1/events/push (ingest external events). SDK-first design: workflows are defined in code and registered with Hatchet via gRPC worker connection; the REST API handles the orchestration plane. Response payloads are JSON with run status, step results, and timing metadata. Server-Sent Events (SSE) available for real-time run progress streaming.

API token auth via Authorization: Bearer header. Tokens are per-tenant and scoped to the Hatchet instance. Tenant tokens created via Hatchet dashboard or CLI. No OAuth2 for direct API access. Worker registration uses the same API token with gRPC connection to the engine. HTTPS enforced on Hatchet Cloud; self-hosted instances require operator TLS configuration.

Structured JSON error responses with code and message fields. Failed workflow steps surface error details in run status payload. Dead-letter queue behavior configurable per workflow step (max retries, backoff). Worker health is monitored by the engine; unhealthy workers trigger step reassignment. Hatchet Cloud uptime tracked on status.hatchet.run. Self-hosted reliability depends on operator infrastructure.

docs.hatchet.run covers quickstart guides for TypeScript, Python, and Go; workflow definition patterns; API reference; self-hosting setup; and Hatchet Cloud onboarding. Documentation is actively maintained alongside the open-source releases. Getting started requires a Hatchet instance (Cloud or self-hosted) and API token. Community support via Discord.