Jetty documentation
Jetty is where you find, share, and run reliable AI workflows. You write a runbook — one markdown file that gives an agent its job, its bar for “done”, and its checks — and Jetty runs it in a managed sandbox, records every run as a trajectory, and helps you improve it until it works every time.
These docs are written for builders: developers who live in Claude Code, Codex, Cursor, or Gemini CLI and want a runbook running in minutes instead of a framework to stand up. Pick a starting point.
Start here
From zero to a runbook you trust, in one sitting.
Get a token, install the skill, and run your first runbook in under ten minutes.
Browse runbooks other builders have published and open one to start from.
Connect your tools
Jetty meets you where you already work. The fastest path is the Claude Code plugin; any MCP-compatible editor works through the MCP server; and TypeScript apps use the client SDK.
A decision matrix across the five ways to connect Jetty to your tools.
Install the plugin, run /jetty-setup, and drive Jetty with /jetty, create-runbook, optimize-runbook.
npx -y jetty-mcp-server — 16+ tools across Cursor, VS Code, Windsurf, Zed, Codex, Gemini CLI.
@jetty/sdk for TypeScript: JettyClient, runAndWait, and the run → check → fix → rerun loop.
Bring your own agent
Already have an agent in another framework? Keep it. Jetty becomes the independent evaluation, observability, and optimization layer around it, or you run one of the four built-in runtimes inside Jetty's own sandboxes.
Keep your agent in any framework; use Jetty as the independent grader, store, and A/B harness. Flue worked example.
The four agents Jetty runs inside its own sandboxes: claude-code, codex, gemini-cli, hermes.
Go deeper
The mental model and the task-shaped guides, when you want them.
The hero artifact: one markdown file with the job, the bar for “done”, and the checks.
How multi-step workflows, the step library, and path expressions move data between steps.
Every run is a trajectory. Evals turn runs into a hill-climbing loop with /optimize-runbook.
One API, two modes (passthrough proxy and runbook sandbox), durable execution, object storage.
The canonical structure: frontmatter, objective, output manifest, evals, and the /create-runbook wizard.
Add evals to a runbook, read the trajectories, and hill-climb with /optimize-runbook.
Run a runbook on a cron cadence so evals stay fresh as models and providers drift.
Trigger workflows from GitHub Actions with eval-driven quality gates.
Provider keys, environment variables, and OAuth credential forwarding into sandboxed runs.
Reference
The chat-completions endpoint (two modes), routines, webhooks, and GitHub-PR APIs.
The catalog of activities: AI models, control flow, data processing, evaluation.
Copy-paste workflow JSON and agent recipes for common tasks.
Common failure modes — parameter mismatches, token handling, runtime quirks — and fixes.
Ready to build?
Start the quickstart → — token to first runbook in under ten minutes.
Prefer to talk it through? Book a 20-minute walkthrough → Bring a task from your own work and we'll turn it into a runbook together.