Durable orchestration for AI agents

Build, observe, and scale AI agents and workflows in a single, easy-to-use platform.

Start for free Docs

Running Billions of Tasks for Scale-Ups and Enterprises

Core Principles

Performance

Built for low-latency, high-throughput workloads with task start times of less than 20ms. Smart assignment rules handle rate limits, fairness, and priorities without complex configuration.

Durability

Every task invocation is durably logged to a data store. When jobs fail, resume exactly where you left off — no lost work, no duplicate LLM calls, no engineering headaches.

Code-First

Hatchet SDKs are language-native so developers can write business logic as versionable, reusable, testable atomic functions.

Use Case

Ingestion & Indexing Pipelines

Agents and AI workflows need continuous, fast access to your data to construct context on-demand. With Hatchet, your vector databases and knowledge graphs will always be up-to-date.

Learn more

Scale with Hatchet

Build resilient pipelines that handle real-world complexity. Automatic retries, intelligent rate limiting, and checkpoint recovery mean your data stays fresh without constant firefighting.

Build RAG, document processing, and indexing pipelines with ease
Easily replay failed pipelines from the Hatchet UI
Update vector databases in real-time with exactly-once semantics

With Hatchet, we've scaled our indexing workflows effortlessly, reducing failed runs by 50% and doubling our user base in just two weeks!

Soohoon ChoiCo-founder · Greptile

Case Study

document-indexing

Loading syntax highlighting...

Use Case

AI Requests & Agents

AI agents need complex orchestration — managing tool calls, handling timeouts, maintaining conversation state, and enforcing safety constraints. Most teams end up building fragile in-process systems that are difficult to scale and maintain.

Learn more

Reliability with Hatchet

Define agents as simple, durable functions with built-in orchestration primitives. Set guardrails, manage state, and handle failures gracefully.

Write your agent's tools as simple functions, integrated tightly with your business logic
Designed to be long-running with safety and security constraints
Built-in eventing for human-in-the-loop signaling and streaming responses

Implementing Hatchet has revolutionized our task management system, enabling us to handle a growing number of background tasks efficiently.

Shaun BerrymanStaff SWE · Moonhub

Case Study

ai-agent

Loading syntax highlighting...

Use Case

Massive-Parallelization

Processing thousands of documents, enriching large datasets, running agent-swarms, or scheduling GPU workloads requires complex coordination. Most solutions either can't scale or become impossibly complex to manage.

Learn more

Parallelize with Hatchet

Fan out to thousands of workers with a single function call. Built-in fairness algorithms and resource management ensure efficient utilization without manual tuning.

Process entire document repositories in parallel
Enrich millions of leads without rate limit headaches
Schedule GPU jobs with intelligent batching
Scrape web data with automatic retry and deduplication

Hatchet enables Aevy to process up to 50,000 documents in under an hour through optimized parallel execution, compared to nearly a week with our previous setup.

Ymir EgilsonCTO · Aevy

Case Study

agent-swarm

Loading syntax highlighting...

How it works

There are two components to running Hatchet: the orchestration engine and your workers. Workers run on your own infrastructure, while the Hatchet orchestration engine is available as a managed service or can be self-hosted.

Step 1
Write tasks and workflows as code

Tasks are simple functions. They can be composed into workflows to represent more complex logic. Your tasks automatically retry on failure and handle complex dependencies between steps.

Step 1
Write tasks and workflows as code

Tasks are simple functions. They can be composed into workflows to represent more complex logic. Your tasks automatically retry on failure and handle complex dependencies between steps.

Step 2
Invoke your tasks and workflows

Start workflows from your API, schedule them to run at specific times, or trigger them when events happen. Tasks run immediately or queue up for later.

Step 2
Invoke your tasks and workflows

Start workflows from your API, schedule them to run at specific times, or trigger them when events happen. Tasks run immediately or queue up for later.

Step 3
Run workers in your own cloud

Deploy workers on Kubernetes, Porter, Railway, Render, ECS, or any container platform. They automatically connect to Hatchet and can scale up or down based on workload.

Step 3
Run workers in your own cloud

Deploy workers on Kubernetes, Porter, Railway, Render, ECS, or any container platform. They automatically connect to Hatchet and can scale up or down based on workload.

Step 4
Monitor and replay

See all your worksflows in the dashboard, get alerts when tasks fail, and export metrics to your monitoring tools. Full visibility (and control) without extra setup.

Step 4
Monitor and replay

See all your worksflows in the dashboard, get alerts when tasks fail, and export metrics to your monitoring tools. Full visibility (and control) without extra setup.

Build AI that scales. Consolidate your legacy orchestration into one reliable, scalable, & secure solution.

Enterprise-grade security, compliance, and SSO.
Processing over 100 million tasks/day for AI-first companies.
Custom deployment options & bring-your-own-cloud available.

Start for free Schedule Enterprise Demo