# Jetty

> Jetty is where professionals 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 work in Claude Code, Codex, or Cursor. Connect via the Claude Code plugin, the MCP server, the TypeScript SDK, or the OpenAI-compatible REST API.

## For agents

- [agent-instructions.md](https://jetty.io/agent-instructions.md): pasteable onboarding trigger — numbered steps, human-only handoffs, ends in a Return block. Start here.
- [MACHINE_CONTEXT.md](https://jetty.io/MACHINE_CONTEXT.md): deep reference — state model, CLI verb contracts, runbook format, anti-patterns.
- [auth.md](https://jetty.io/auth.md): how an agent obtains, uses, and revokes a Jetty API credential. Machine-readable discovery: https://jetty.io/.well-known/oauth-protected-resource, https://jetty.io/.well-known/oauth-authorization-server, https://jetty.io/.well-known/api-catalog.
- [Agent skills index](https://jetty.io/.well-known/agent-skills/index.json): the jetty, jetty-setup, create-runbook and optimize-runbook skills with SHA-256 digests. MCP Server Card at https://jetty.io/.well-known/mcp/server-card.json; ARD capability manifest at https://jetty.io/.well-known/ai-catalog.json.

Every HTML page on this site is also available as Markdown: request it with `Accept: text/markdown`.

- [llms-full.txt](https://jetty.io/llms-full.txt): every docs page below as Markdown, in one file.

## Get Started

From zero to a runbook you trust, in one sitting.

- [Get Started](https://jetty.io/docs/get-started): What Jetty runs, and three ways in: start free, install locally, or call it from your app.
- [Try a Project](https://jetty.io/runbooks): Browse the Jetty Runbook Directory and try it yourself on your own account.
- [Bring Your Agent](https://jetty.io/docs/integrations/bring-your-own-agent): Install Jetty locally: the plugin, the jetty-setup skill, and your first run.
- [Connect Your App](https://jetty.io/docs/get-started/connect-your-app): Every on-ramp: the MCP server, the Client SDK, and the API.

## Product Overview

What Jetty does, in two motions: run the workload, then read the run.

- [Product Overview](https://jetty.io/docs/product-overview): The two halves of Jetty: run AI workloads in managed sandboxes, and investigate how well the agent performed.
- [Running AI Workloads](https://jetty.io/docs/product-overview/running-ai-workloads): How a runbook becomes a running workflow: the author → run → investigate → improve lifecycle, and what Jetty assembles around every run.
- [Investigating Agent Quality](https://jetty.io/docs/product-overview/investigating-agent-quality): How to read a run and its verdict, add a check at the failure moment, compare runs honestly, and run the investigate-and-improve loop.

## AI Workloads

The job you hand to Jetty, and the pieces that carry it out.

- [AI Workloads](https://jetty.io/docs/ai-workloads): What counts as an AI workload, and the two pieces every one is built from: a runbook and a runtime.
- [Runbooks](https://jetty.io/docs/ai-workloads/runbooks): The hero artifact: one markdown file with the job, the bar for “done”, and the checks.
- [Runtime Configuration](https://jetty.io/docs/ai-workloads/runtime-configuration): The agents Jetty runs inside its own sandboxes, which providers back each, and how to pick one.
- [antigravity](https://jetty.io/docs/ai-workloads/runtime-configuration/antigravity): Google's current terminal agent (agy) in Gemini-API-key mode: catalog slugs, effort tiers, MCP support.
- [claude-code](https://jetty.io/docs/ai-workloads/runtime-configuration/claude-code): The default runtime: Anthropic's agent with MCP, full usage telemetry, and the loop Jetty's skills are tuned against.
- [codex](https://jetty.io/docs/ai-workloads/runtime-configuration/codex): OpenAI's agent on a pinned release: Responses-API only, config.toml routing, the reasoning-effort knob.
- [gemini-cli](https://jetty.io/docs/ai-workloads/runtime-configuration/gemini-cli): Google's earlier terminal agent; kept for existing runbooks; start new Gemini work on antigravity.
- [goose](https://jetty.io/docs/ai-workloads/runtime-configuration/goose): Block's agent, included for comparison: bare model ids, direct providers, best-effort output parsing.
- [hermes](https://jetty.io/docs/ai-workloads/runtime-configuration/hermes): Nous Research's agent; the one runtime that reaches all six providers, including self-hosted vLLM.
- [opencode](https://jetty.io/docs/ai-workloads/runtime-configuration/opencode): The open-source SST agent with permissions pre-allowed and Playwright + Firecrawl MCP ready on every run.
- [pi](https://jetty.io/docs/ai-workloads/runtime-configuration/pi): A small provider-agnostic agent with no MCP client by design: CLI tools and skills instead.
- [Integrations](https://jetty.io/docs/ai-workloads/integrations): How a workload connects to the world: integrations are environment variables injected into every run: Slack, databases, GitHub, CRMs, your own app.
- [Runs](https://jetty.io/docs/ai-workloads/runs): Every run is captured end to end: each step, its inputs, outputs, files, and status, plus labels to group and compare runs.
- [Agent Evaluations](https://jetty.io/docs/ai-workloads/agent-evaluations): Code checks, checklists, and LLM or agent judges live in the runbook and run in flight. Every run records a verdict you can investigate and improve against.
- [REST API](https://jetty.io/docs/reference/api): The Jetty REST API: run runbooks as tasks (deploy, run, monitor — the primary endpoint), the OpenAI-compatible chat-completions compatibility layer, schedules, webhooks, and GitHub-PR APIs.
- [Machine Instructions](https://jetty.io/docs/reference/api/machine-instructions): The runbook API surface for agents and programmatic callers: tasks, runs, init_params, trajectories, files, MCP config, sweeps, snapshots, webhooks, schedules.
- [Actions](https://jetty.io/docs/reference/api/steps): The pre-built actions that run alongside runbooks — AI models, control flow, data processing, evaluation — and the path expressions that wire them together.

## Jetty and Friends

Run your agent on Jetty or anywhere, and evaluate on Jetty.

- [Jetty and Friends](https://jetty.io/docs/jetty-and-friends): Keep your agent in any framework; use Jetty as the independent grader, store, and A/B harness. Flue and eve worked examples.
- [Bring Your Friends and Framework](https://jetty.io/docs/jetty-and-friends/bring-your-friends-and-framework): Keep your agent in its own framework and let Jetty grade every run: the pattern, with Eve and Flue worked examples.
- [Eve + Jetty](https://jetty.io/docs/jetty-and-friends/bring-your-friends-and-framework/eve): Mount the @jetty/eve extension into a Vercel eve agent: live grading, a grade-steered experiment bandit, and an eval reporter. Catch a regression before it ships.
- [Flue + Jetty](https://jetty.io/docs/jetty-and-friends/bring-your-friends-and-framework/flue): The complete flue-jetty walkthrough: grade every Flue agent run with an independent rubric and catch a regression before it ships.
- [Client SDK](https://jetty.io/docs/integrations/sdk): @jetty/sdk for TypeScript: JettyClient, runAndWait, etc.
- [MCP Server](https://jetty.io/docs/integrations/mcp-server): 27 tools for Claude Code, Cursor, VS Code, Windsurf, Zed, Codex, and WebMCP.

## Guides

Task-shaped walkthroughs for the work you actually do.

- [Guides](https://jetty.io/guides): Every guide in one place: use cases from teams running AI workloads, improving a runbook from its own runs, and scheduling workloads.
- [Build vs. Buy Evaluation](https://jetty.io/docs/guides/ai-agent-evaluation-cost): What AI agent evaluation costs to build versus buy: the variables that matter, a framework for making the call, and how to test it against your own workload.
- [Building AI Evaluations](https://jetty.io/docs/guides/building-ai-evaluations): A way to know, before you ship, whether your agent is actually doing the job it was built to handle at the level of quality you expect.
- [Creating a Benchmark](https://jetty.io/docs/guides/creating-a-benchmark): Pick one axis of comparison, write the runbook that defines the job, and run the task set against every candidate in parallel so the comparison is reproducible.
- [Defining Agent Checklists](https://jetty.io/docs/guides/ai-agent-checklist): Keep your agent on task with a checklist: the conditions that define done, ticked and rechecked inside every run, then investigated across runs and kept honest in production.
- [Improving Agent Runbooks](https://jetty.io/docs/guides/improving-agent-runbooks): Pull the latest runs with your coding agent, tighten the evals they expose, and let the optimize-runbook skill propose evidence-backed edits.
- [Research](https://jetty.io/guides/research): Papers Jetty has contributed to, on agent benchmark standards, reproducible ML evaluations and AI accountability.
- [Scheduling AI Workloads](https://jetty.io/docs/guides/scheduling-ai-workloads): Run a runbook on a cron cadence so evals stay fresh as models and providers drift.
- [Use Cases](https://jetty.io/guides/use-cases): Stories from teams running AI workloads on Jetty: what they set up, how parallel runs and evals changed the work, and what came out.

- [Index](https://jetty.io/docs/overview): What Jetty is, who it is for, and every docs page in one place.

## Source

- [jettyio/jettyio-skills](https://github.com/jettyio/jettyio-skills): Claude Code plugin, skills, and the MCP server.
- [jettyio/jetty-sdk](https://github.com/jettyio/jetty-sdk): the TypeScript client SDK and the Flue integration example.