---
title: "Agent Fleet · Multi-agent orchestration for the Claude Code"
description: "Multi-agent orchestration for the Claude Code CLI. Run specialized AI agents in parallel: research, critique, analyze, fix, and discuss. TypeScript"
canonical: "https://matthiasmeyer.tech/repos/agent-fleet"
language: "en"
last_updated: "2026-08-19"
markdown_versions: ["https://matthiasmeyer.tech/repos/agent-fleet.md", "https://matthiasmeyer.tech/de/repos/agent-fleet.md", "https://matthiasmeyer.tech/es/repos/agent-fleet.md"]
image: "https://matthiasmeyer.tech/api/og/repo/agent-fleet?lang=en"
publisher: "Matthias Meyer, https://matthiasmeyer.tech (llms.txt: https://matthiasmeyer.tech/llms.txt)"
---

agents stable TypeScript

# Agent Fleet

Multi-agent orchestration for the Claude Code CLI

Run specialized AI agents in parallel: research, critique, analyze, fix, and discuss. TypeScript orchestration layer.

[View on GitHub →](https://github.com/studiomeyer-io/agent-fleet)[npm](https://www.npmjs.com/package/@studiomeyer-io/agent-fleet)

## What it solves

Most agent frameworks treat a language model as an API endpoint: you manage the tokens, you hand it tools through function calls, you glue the prompts together. The agent never really touches your machine, so anything it wants to read or change has to be passed through your own plumbing first. Agent Fleet takes the other route and runs each agent as a full Claude Code subprocess, so it reads files, edits them and runs commands the way a developer would.

## How it works

Seven agents ship with it, each with a job: Research does deep research in eight modes, Critic plays devil advocate against a plan, Analyst reads a codebase for patterns and health, Discovery scans for security holes, dead code, type gaps and debt, Repair takes what Discovery found and fixes it, CTO makes live changes during a discussion, and Conductor runs three or more of them in parallel and synthesises what they said. Repair and CTO both have a dry-run switch that keeps them read-only. Since version 0.2 there is an optional LangGraph mode with Postgres checkpoints, so a crashed run resumes and a high-risk finding can pause for a human.

## When to use it

It fits when you want several opinions on the same problem and you want them to actually look at the code, not at a pasted excerpt. It runs on your existing Claude subscription by default, so there is no token counting, and it takes an API key instead if you need that for commercial use.

## When not to use it

Because the agents are real subprocesses with real file access, they are as dangerous as the directory you start them in. Repair and CTO write to disk unless you pass the dry-run switch. It also needs the Claude Code CLI installed, which makes it a developer-machine tool rather than something you embed in a product.

Stars

★ 1

Forks

⑂ 1

installs

↓ 256

Clones (14d)

⊟ 60

Language

TypeScript

Updated

2026-08-19

## Install

```
npx -y @studiomeyer-io/agent-fleet
```

## Related repos

- MCP OTel

    ← distributed tracing for

    →

    https://matthiasmeyer.tech/repos/mcp-otel

- StatePilot

    ← guards the tool flow of

    →

    https://matthiasmeyer.tech/repos/statepilot

- SkillDoctor

    ← lints the skill files of

    →

    https://matthiasmeyer.tech/repos/skilldoctor

- Local Memory MCP

    → stores context in

    →

    https://matthiasmeyer.tech/repos/local-memory-mcp

- Personal Suite MCP

    → calls tools from

    →

    https://matthiasmeyer.tech/repos/mcp-personal-suite

- AI Shield

    ← shields

    →

    https://matthiasmeyer.tech/repos/ai-shield

- Darwin Agents

    ← evolves prompts in

    →

    https://matthiasmeyer.tech/repos/darwin-agents

- MCP Hook Conformance

    ← validates hooks for

    →

    https://matthiasmeyer.tech/repos/mcp-hook-conformance

- Darwin LangGraph

    ← alt. orchestration for

    →

    https://matthiasmeyer.tech/repos/darwin-langgraph

## Sitemap

Every page of this site as Markdown: [sitemap](https://matthiasmeyer.tech/sitemap.md). Curated entry point: [llms.txt](https://matthiasmeyer.tech/llms.txt).