Documentation

Everything you need to run a mesh.

Start with the Quickstart to get a team running, then come back for the details. Every page is on this site, so search engines and your own browser find it without a trip to a wiki.

Source on GitHubPackage on npm

Getting started

Install ClaudeLink and run your first coordinated team.

QuickstartOne command, three terminals, and a working mesh in about a minute.InstallationPer-CLI setup for Claude Code, Codex, Gemini, and Goose, global or per project.

Running the mesh

The pieces you use day to day once agents are connected.

Command CenterThe live local console at 127.0.0.1:7878: panels, lifecycle, and how to opt out.Auto-nudgeThe scheduler and the Claude Code Stop hook that make collaboration hands-free.Tools referenceEvery MCP tool an agent gets: register, send, broadcast, read_inbox, and the rest.CLI referenceEvery claudelink command: init, status, ui, install-hooks, reset.ConfigurationWhere state lives, per-CLI config files, and the environment knobs.

Understanding it

How ClaudeLink works under the hood, and what it does and does not touch.

ArchitectureThe MCP layer, the SQLite mesh, message flow, and the agent lifecycle.Security and privacyLocal-first by design: no cloud, no telemetry, loopback-only, MIT licensed.

Reference

Answers to common questions and where the project is headed.

FAQDisabling it, Windows support, adding a new CLI, and other common questions.RoadmapMulti-machine support, more CLI integrations, channels, and search.