Hatchet Mcp
Let an agent list Hatchet workflow runs, read logs, trigger, cancel, and replay jobs without leaving the editor during on-call or release debugging.
Overview
Hatchet MCP is an MCP server for the Operate phase that lets an AI agent observe and control Hatchet workflow runs, logs, triggers, cancellations, and replays.
What is this MCP server?
- Inspect workflow runs and logs from the agent
- Trigger, cancel, and replay Hatchet workflows via MCP tools
- stdio npm package hatchet-mcp (v0.1.1)
- Configurable via HATCHET_CLIENT_TOKEN JWT; optional HATCHET_API_BASE for self-hosted instances
- npm package hatchet-mcp v0.1.1
- Primary auth: HATCHET_CLIENT_TOKEN (required secret)
Community signal: 1 GitHub stars.
What problem does it solve?
Debugging stuck or failed Hatchet runs forces context switches between the IDE, dashboard, and CLI while production jobs keep failing.
Who is it for?
Solo builders on Hatchet who want agent-assisted incident response and routine job operations without leaving the editor.
Skip if: Teams not using Hatchet, or greenfield projects still choosing a workflow engine with no token yet.
What do I get? / Deliverables
Once hatchet-mcp is installed with a valid token, the agent can query runs, pull logs, and trigger or replay workflows from your normal coding session.
- Agent-driven visibility into Hatchet run status and logs
- Ability to trigger, cancel, or replay workflows from the IDE
Recommended MCP Servers
Journey fit
Workflow observation and control matter after you ship durable background jobs—canonical shelf is Operate, not first-time feature coding. Hatchet MCP targets running infrastructure and job orchestration, which aligns with operate → infra rather than frontend build work.
How it compares
Hatchet orchestration MCP bridge, not a generic cron skill or Terraform deploy skill.
Common Questions / FAQ
Who is Hatchet MCP for?
Developers already running Hatchet workflows who want Claude Code, Cursor, or similar agents to manage runs and logs.
When should I use Hatchet MCP?
Use it in operate when jobs fail, need replay, or you want the agent to trigger workflows while fixing backend or infra code.
How do I add Hatchet MCP to my agent?
Install the npm package hatchet-mcp, set HATCHET_CLIENT_TOKEN (required JWT), add the stdio server to your MCP config, and optionally set HATCHET_API_BASE for self-hosted APIs.