hatchet

Why we included this project

Hatchet sits one level above a plain job queue: background work runs as durable steps that can be replayed after a crash instead of vanishing. That matters for agent loops and long-running pipelines, where a failed step should trigger a retry rather than force a full restart. The engine handles DAG pipelines, scheduled runs, and event-triggered work, with first-party SDKs for Python, TypeScript, Go, and Ruby, so one platform can carry both an agent loop and a batch backfill. You can use the managed cloud or self-host on your own Postgres-backed stack, which gives teams control over where workloads execute. The built-in web UI and alerting, plus OpenTelemetry tracing, make it a practical fit for production systems where silent job failure is not an option.