Welcome to Jetty!

Jetty is a platform for running AI agents and workflows in production.

Three Use Cases

1. Agentic Workflows

Describe outcomes in English. Jetty provisions a sandbox, runs the agent, persists artifacts, returns results. No infrastructure to manage.

Learn more →

2. Workflow Orchestration

Build multi-step AI data pipelines with built-in evaluation. Chain LLM calls, agent runs, quality gates, and control flow into DAGs.

Learn more →

3. Jetty Agent

Connect your observability telemetry. Jetty analyzes your LLM usage and generates pull requests with optimizations — human-reviewed before anything ships.

Learn more →

Getting Started

Choose your path based on what you need:

Quick Start

• Quickstart: Agentic Workflows — Run an AI agent in a sandbox in 5 minutes
• Setup — Environment setup and API keys
• Your First Flow — Build a workflow step by step

Guides

• CI Integration — Trigger Jetty workflows from GitHub Actions
• Langfuse Setup — Connect telemetry for automated analysis
• Evaluating LLMs — Build evaluation pipelines
• All Guides →

API Reference

• Chat Completions — One endpoint, two modes (LLM passthrough + runbook execution)
• Webhook API — Receive workflow completion callbacks
• GitHub PR API — Create pull requests programmatically
• Full API Reference →

For AI Coding Agents

• Agent & MCP Overview — Three ways to connect (plugin, MCP, REST)
• Claude Code Plugin — Install and use /jetty commands
• MCP Server — Setup for Cursor, Windsurf, VS Code, Zed, and more

Quick Example

{
  "model": "claude-sonnet-4-6",
  "messages": [
    {"role": "system", "content": "Analyze the uploaded CSV and produce a report."},
    {"role": "user", "content": "Run the analysis"}
  ],
  "stream": true,
  "jetty": {
    "runbook": true,
    "collection": "my-org",
    "task": "analyze-data",
    "agent": "claude-code",
    "file_paths": ["uploads/dataset.csv"]
  }
}

Without the jetty block — standard LLM passthrough (100+ providers). With it — full agent sandbox execution.