Hatchet
Hatchet is an open-source distributed task queue and workflow orchestration engine for background jobs, AI agents, and durable workflows. It is Postgres-backed, MIT-licensed, and ships with first-class SDKs for Python, TypeScript, Go, and Ruby plus a managed offering (Hatchet Cloud) and a self-hostable engine (Docker Compose, Hatchet Lite, Helm chart). The platform separates the orchestration engine from worker execution so workers can run on the operator's own infrastructure (Kubernetes, ECS, Render, Railway, Porter, or any container platform).
APIs
Hatchet API
The Hatchet REST API is the control plane for the Hatchet engine. It exposes operations for tasks, workflow runs, durable tasks, events, filters (CEL-based event routing), webho...
Collections
Hatchet API
OPENPricing Plans
Rate Limits
FinOps
Hatchet Finops
FINOPSFeatures
Task and workflow state is persisted to Postgres so executions survive worker restarts, network partitions, and engine upgrades.
Sub-20ms task start times with intelligent assignment rules.
Native language SDKs for Python, TypeScript, Go, and Ruby; tasks are versionable, testable functions.
The orchestration engine and workers are decoupled; workers run on the customer's own infrastructure.
Hatchet's only hard dependency is PostgreSQL; RabbitMQ is optional for higher-throughput deployments.
Engine-level rate-limit primitive throttles task execution by named key, scope, and window — independent of HTTP rate limits.
Common Expression Language filters bind incoming events to workflows with payload predicates.
One-shot scheduled runs and recurring cron-triggered workflows are first-class engine resources.
External HTTPS endpoints can be registered as workers; the engine delivers task runs as signed webhook requests.
Workflows and tasks emit OTel traces; the engine exposes a trace lookup endpoint.
Failed or cancelled tasks can be replayed with the same input or restored to the run queue.
Tenants isolate workflows, tasks, workers, and tokens; quotas and throughput are scoped per tenant.
Built-in web dashboard for inspecting runs, replaying tasks, and managing tenants.
Alerts can be dispatched to Slack channels or AWS SNS topics for ops escalation.
Engine, API, SDKs, dashboard, and Helm chart are all MIT-licensed.
Use Cases
Manage tool calls, conversation state, timeouts, and checkpointing for production AI agents.
Classic distributed task-queue use case — replace Celery, Sidekiq, BullMQ, or RQ with durable equivalents.
Multi-step pipelines with retries, conditional branching, and exactly-once semantics.
Keep vector databases, knowledge graphs, and search indexes up-to-date as upstream sources change.
Fan out to thousands of workers for batch processing and embarrassingly parallel workloads.
Workloads where retries, checkpointing, and replay are non-negotiable (payments, billing, compliance).
Use events + CEL filters to route external signals into the right workflows without bespoke routing code.
Integrations
The durability substrate; Hatchet ships with managed migrations and is happy on RDS, Cloud SQL, Neon, Supabase, or self-hosted Postgres.
Optional message bus for inter-service communication and high-throughput real-time updates.
Official Helm chart at hatchet-dev/hatchet-charts for production self-hosting.
Hatchet Lite single-image deployment plus a production Docker Compose stack.
ECS-friendly worker deployment patterns and first-class SNS alerting.
Native Slack alerting integration for run failures and SLA breaches.
GitHub OAuth login for the dashboard plus repo-linked workflow source.
Engine and SDKs emit OTel traces; the API exposes trace lookup.
Official Terraform provider for managing Hatchet Cloud resources as code.
brew install via the hatchet-dev/homebrew-hatchet tap.
Documented quickstart for combining Hatchet Python SDK with FastAPI services.
Multiple Next.js starter templates wire Hatchet into App Router and Pages Router projects.
Hatchet's durable-execution model is a natural substrate for orchestrating Anthropic Claude tool calls and multi-step agents; the hatchet-typescript-deep-research and hatchet-typescript-code-agent reference patterns illustrate the integration shape.
Solutions
Fully managed orchestration engine on Hatchet's infrastructure with tier-included task-run allowances and SOC 2 / HIPAA controls.
MIT-licensed engine, API, dashboard, and Helm chart for running Hatchet on the operator's own infrastructure.
Single-image bundled deployment of engine, API, and dashboard for development, testing, and low-throughput production.
Enterprise tier deploys the Hatchet engine inside the customer's own VPC while remaining managed by Hatchet.