Hatchet
hatchet.run
- Free plan
- Linux
- Web
- API
- Self-hosted
What Hatchet is
Hatchet is a developer platform for building and deploying AI agents, durable workflows, and background tasks. It saves tasks and agent invocations so teams can inspect executions, retry work, replay runs, and build workflows that persist. Workflow controls include scheduled and cron runs, event and webhook triggers, concurrency settings, priorities, and rate limits. SDKs are documented for Python, TypeScript, Go, and Ruby. For monitoring, Hatchet provides real-time workflow and task views, an OpenTelemetry collector, and searchable application logs. The orchestration engine can run on Hatchet Cloud or be self-hosted; workers run on the customer's infrastructure. Local development and CI can use embedded mode without an account or Docker, with CLI, Docker Compose, and Kubernetes options also listed. The Pay-as-you-go plan is free to start and includes 1 million runs and 1 million events monthly; excess usage is billed at $30 per million runs and $5 per million events. SOC 2 Type II is included on every listed tier.
Who it suits
Hatchet says it is particularly suited to AI agents, massive parallelization, and mission-critical workloads. Its deployment choices and SDKs may suit teams building and operating workflows across different environments.
Good to know, and what to watch
- Retries, schedules, event triggers, priorities, and rate limits
- SDKs for Python, TypeScript, Go, and Ruby
- Live monitoring, OpenTelemetry collector, and searchable logs
- Embedded local mode requires no account or Docker
- Free plan includes 1 million runs and events monthly
- Free-tier limits can cause tasks to be dropped
- Usage above included allowances is billed separately
Plans
Every plan and priceCustom
Pay-as-you-go
Free plan · paid from $10
Getting started
- Open https://hatchet.run/
- Choose Hatchet Cloud or a self-hosted orchestration deployment
- Select an SDK for Python, TypeScript, Go, or Ruby
- Run workers on your own infrastructure
- Use embedded mode for local development or CI without an account or Docker
- Deploy with the CLI, Docker Compose, or Kubernetes
Limits to check
The free Pay-as-you-go allowance covers 1 million runs and 1 million events per month, up to 100 workers, 10 tenants, and 50 users. It includes seven-day retention and 10 GB per day of active storage; above included usage, billing applies, and exceeded free limits can cause dropped tasks or errors creating new tasks.
The Club's review
Hatchet is a developer platform for running AI agents, durable workflows, and background jobs. It fits teams that need high parallelism or business-critical processing while keeping workers on their own infrastructure. Its durable execution and operational controls are compelling, but teams must plan around hard free-tier caps and metered usage.
Overview
Hatchet stores tasks and agent invocations durably, making failures inspectable and executions retryable, replayable, and resumable. That is a meaningful advantage over job systems where an interrupted run simply disappears. The orchestration engine can run in Hatchet Cloud or be self-hosted, while workers run on infrastructure the customer controls, giving teams a hybrid deployment model.
Embedded mode supports local development and CI without an account or Docker. CLI, Docker Compose, and Kubernetes options provide a path from a local experiment to a managed deployment. SDKs are documented for Python, TypeScript, Go, and Ruby, so adoption does not require standardizing a single application language.
Key features
Durable workflow execution
Automatic retries, scheduled and cron runs, event and webhook triggers, priorities, concurrency controls, and rate limits cover the practical demands of production background processing. Concurrency and rate limits help protect downstream APIs and databases; priorities let urgent work move ahead of routine jobs. Human approvals are supported when a workflow needs an explicit checkpoint rather than fully autonomous execution.
Monitoring and integrations
Real-time workflow and task monitoring, searchable application logs, and a built-in OpenTelemetry collector give operators a way to follow execution and investigate failures. Hatchet says it works with Claude Code and Cursor, and lists Kubernetes, Docker, ECS, Cloud Run, Porter, and Railway as deployment options. These integrations are useful for teams already using those environments, while workers remaining on customer infrastructure may simplify network and data-placement decisions.
Security and enterprise controls
SOC 2 Type II is included on every listed tier. The Custom plan adds SSO/SAML, RBAC, audit logs, and a HIPAA BAA, which are the controls regulated or centrally administered organizations are likely to require. Dedicated Slack support is listed for Custom; the Pay-as-you-go comparison includes Discord, while its plan description specifies community support.
Pricing
Pay-as-you-go
The free tier starts at no cost and includes 1 million runs and 1 million events per month. Usage above those allowances is billed at the end of each 30- or 31-day cycle: $30 per million runs and $5 per million events. The plan allows up to 100 workers, 10 tenants, and 50 users, with seven-day retention and 10 GB/day of active storage. It suits teams validating an agent or workflow platform and workloads that stay within those quotas.
The cap behavior matters more than the headline price: when a free-tier limit is exceeded, tasks are dropped and attempts to create new tasks return errors. That is a poor fit for workloads that must continue through a usage spike unless billing and quota monitoring are in place.
Custom
Custom pricing is aimed at organizations running 100 million or more runs per month, requiring dedicated capacity, or needing enterprise features. It adds volume discounts, latency guarantees and SLAs, multi-region operation, SSO, RBAC, audit logs, HIPAA support, and self-hosted and BYOC support. Buyers with strict availability, compliance, or geographic requirements should discuss this tier; smaller teams are unlikely to justify its sales-led model.
Platforms
Hatchet is offered through an API and web interface, supports Linux, and can be self-hosted. The cloud-or-self-hosted engine plus customer-managed workers is a practical choice for organizations balancing managed orchestration with infrastructure control.
Who it's for
Hatchet is particularly suited to AI-agent systems, massive parallelization, and mission-critical workloads where replayable execution and operational visibility matter. It is also a strong candidate for polyglot engineering groups and teams that need to deploy workers in Kubernetes, containers, or other infrastructure they already operate.
It is less suitable for a small, low-volume queue that only needs simple delayed jobs, or for teams unable to tolerate dropped work when a free quota is crossed. The usage model rewards capacity planning rather than casual, unmonitored adoption.
Pros and cons
Pros
- Durable execution records: retries, debugging, replay, and workflow resumption make failures recoverable.
- Strong production controls: schedules, triggers, priorities, concurrency, rate limits, and human approvals address complex operating conditions.
- Flexible deployment: Hatchet Cloud or self-hosted orchestration with workers on your infrastructure supports control-sensitive environments.
- Broad developer reach: Python, TypeScript, Go, and Ruby SDKs, plus embedded local mode, reduce adoption friction.
- Useful observability: real-time monitoring, searchable logs, and OpenTelemetry help operators diagnose running systems.
Cons
- Hard free-tier failure mode: over-quota tasks are dropped and new task creation errors, creating operational risk without active usage controls.
- Metered overage: run and event charges can grow with high-volume agent activity.
- Enterprise features are sales-led: SSO, RBAC, audit logs, HIPAA, multi-region, and SLAs require the Custom conversation.
- Retention is limited on Pay-as-you-go: seven days may be short for investigations that need a longer history.
Alternatives
For a broader category comparison, see Task Queue Software, Process Orchestration Software, and Workflow Orchestration Software.
- IronWorker is worth choosing when a free public-cluster plan with one concurrency, five hours per month, and 512 MB RAM is enough, or when Windows support matters.
- Faktory suits teams wanting an AGPL open-source core, with enterprise packs priced at $99 a month.
- Google Tasks is the simpler free option for personal lists, subtasks, due dates, and notifications across Android, iOS, and web.
- Graphile Worker fits buyers seeking a free MIT-licensed worker or a sponsorship-based Pro launch offer from $25+ per month.
- JobRunr is an alternative for teams wanting a free OSS plan with recurring-job capacity and API or self-hosted deployment.
- River is better when reliable background jobs and a web UI are enough and workflows or concurrency limits are unnecessary; its Open Source plan has no support.
- Upstash QStash fits API-centric workloads that can stay within 1,000 messages per day and 50 GB bandwidth on its free plan.
- Hangfire is a free, commercially usable option for teams targeting its SQL Server or in-memory storage model on supported platforms.
Verdict
Choose Hatchet when durable, observable execution for AI agents or large parallel workloads is more important than a minimal queueing setup, especially if workers must remain on your infrastructure. Its main reason to look elsewhere is the free tier's abrupt over-quota behavior and the prospect of usage charges; simpler jobs or tightly capped budgets may be better served by a lighter alternative.
The verdict
Hatchet combines durable workflow execution, developer controls, and monitoring for agent and background-task workloads. Teams should account for the free-tier limits and usage charges, or ask about the Custom plan for enterprise features.
Questions people ask
Is Hatchet free?
Yes. The Pay-as-you-go plan starts at $0 and includes 1 million runs and 1 million events per month; usage above included amounts is billed.
Which deployment models are available?
Hatchet Cloud and self-hosted orchestration are available, with workers running on customer infrastructure.
Which languages have SDKs?
Documented SDKs are available for Python, TypeScript, Go, and Ruby.
Does Hatchet support event triggers and retries?
Yes. It supports event and webhook triggers, plus automatic retries.
What enterprise controls are listed?
The Custom plan lists SSO/SAML, RBAC, audit logs, HIPAA BAA, multi-region, and dedicated capacity.
What the maker publishes
- Company name
- The pricing page footer identifies the company as Hatchet Technologies. source
- Deployment
- The orchestration engine is available as Hatchet Cloud or self-hosted, while workers run on the customer's own infrastructure. source
- Durable execution
- Hatchet durably persists tasks and agent invocations for debugging, retries, replays, and durable workflows. source
- Enterprise controls
- The Custom plan lists SSO/SAML, RBAC, audit logs, and HIPAA BAA as included features. source
- Founded
- 2024 source
- Headquarters
- United States source
- Integrations
- Hatchet says it works with Claude Code and Cursor and lists Kubernetes, Docker, ECS, Cloud Run, Porter, and Railway as deployment options. source
- Local development
- Hatchet offers embedded mode for local development and CI without an account or Docker, as well as CLI, Docker Compose, and Kubernetes options. source
- Observability
- Hatchet provides real-time workflow and task monitoring, a built-in OpenTelemetry collector, and searchable application logs. source
- Product
- Hatchet is a developer platform for building and deploying AI agents, durable workflows, and background tasks. source
- SDK languages
- Hatchet documents SDKs for Python, TypeScript, Go, and Ruby. source
- Security
- SOC 2 Type II is included on every listed pricing tier. source
- Support
- The Pay-as-you-go plan lists community support; the comparison lists Discord for both Pay-as-you-go tiers and dedicated Slack for Custom. source
- Usage limits
- If a user exceeds a free-tier limit, Hatchet says tasks will be dropped and attempts to create new tasks will return errors. source
- Who it is for
- Hatchet says it is particularly suited to AI agents, massive parallelization, and mission-critical workloads. source
- Workflow features
- Hatchet supports automatic retries, scheduled and cron runs, event and webhook triggers, concurrency controls, priorities, and rate limits. source
Checked against the maker's pages on 7 Oct 2026.