Labsco
MCP SERVER · OFFICIAL PROJECT

Jetty.io

by jettyio

Find, run and grade Jetty runbooks from the editor you already work in.

Reasoning Scaffolds & Agent Workflow EnginesOfficial source
Summary
The whole runbook platform, as tools in whichever editor you use.

The same 16 core tools back the Claude Code plugin and every other editor, so an agent can go from listing what exists to creating, running and inspecting a runbook without you leaving the chat. It runs as a local process rather than a hosted endpoint, so the token that calls the API stays on your machine.

What it is

A local server from Jetty that exposes the runbook platform — collections, tasks, workflow runs, trajectories and routines — as tools your editor's agent can call. Your editor launches it with npx, and it talks to the Jetty API on your behalf using your token.

What you get
  • Collections and tasks you can list, read, create and update from the conversation
  • Workflow runs started in the background, or run to completion in one call when you want the result inline
  • Trajectories, their stats, and labels you add — so a finished run can be inspected and graded
  • The step-template library, for assembling a task out of steps that already exist
  • A check on which secrets are set, and environment variables written without opening the dashboard
  • Routines: list them, create one, or run one immediately
Requirements

A Jetty API token, read from JETTY_API_TOKEN or a file in your config directory, and npx to launch the server. JETTY_API_URL points it at a non-default backend.

Setup effort

One command — uvx mlcbakery