Product
ClaudeLink
An open-source server that turns the AI coding agents you already run into one coordinated team.
What it is
ClaudeLink is an open-source MCP server that connects multiple AI coding agents into a single team. Open a few terminals, give each one a role, and they share a real-time message bus, a bulletin board, and a hands-free loop that keeps them working together when you step away.
The name says Claude, but the mesh is model-agnostic. It speaks MCP, an open protocol, so Claude Code, Codex CLI, Gemini CLI, Goose, and any other compliant client can join, and each one chooses its own backing model. ClaudeLink is the connective tissue between agents, not a model or a cloud service of its own.
Built for
Developers who would otherwise spend days building agent orchestration plumbing. People running several coding agents at once who are tired of typing "check messages" into every terminal. Anyone who wants to direct a small team of agents on one project without sending their code to someone else's cloud.
Why it exists
A single agent works in sequence. A team of agents that talk to each other, review each other, and keep working while you think is a different kind of leverage. The blocker has always been the person in the middle relaying messages by hand. ClaudeLink removes that step so one operator can run a whole team.
How it is different
Coordination is the product
ClaudeLink does not run models or write code itself. It connects the agents you already run so they can message each other, share a bulletin board, and keep working without you relaying every update by hand.
Any model, any client
The mesh speaks MCP, an open protocol, so any compliant client can join and each one picks its own backing model. A Claude reviewer, a Codex developer, and a Gemini tester on one mesh is a fully supported pattern.
Local first, MIT licensed
Every byte of state lives in one SQLite file on your machine. No cloud, no telemetry, no account. The whole project is open source, so you can read every line and run a custom build.
What you can do with it
A heterogeneous agent mesh
Each terminal connects to one shared message bus and joins a single team, whatever model is on the other end.
- Claude Code, Codex CLI, Gemini CLI, and Goose install with one command each
- Any other MCP-compatible client can join by pointing its config at the server
- Agents send direct messages, broadcast to the group, and post to a persistent bulletin board
Hands-free auto-nudge
A scheduler wakes any agent that has unread mail, so the team keeps moving after you step away.
- Interval is yours to set, from 1 to 120 minutes
- It only nudges terminals that actually have unread mail, so idle agents never burn a turn
- Works across iTerm2 and tmux through their public scripting APIs, with no accessibility shims
A live Command Center
A local web console shows every agent as a peer: role, message counts, last-seen, and a per-agent auto-reply toggle.
- Binds to 127.0.0.1 only, so it is never reachable from outside your machine
- Launches with the first agent and survives restarts
- A recent-messages feed and a one-click way to drop the whole mesh
Durable shared state
All coordination flows through a single SQLite database in WAL mode, safe for concurrent agents.
- No daemon and no background service; each agent spawns its own short-lived server
- Messages and the bulletin board persist across restarts, so no session is amnesiac to the others
- A crash never loses data
Use cases
Run a team while you are away
Keep one agent as your point of contact and let it relay between you and the rest of the team. Approve a decision through one channel and the mesh carries it forward.
Compress the timeline
One terminal fans work out to several developer terminals, reviews what comes back, and spins up auditors. Work that would take one agent days, run side by side, can land in a few hours.
Many agents, one repository
Put an architect, a couple of developers, and a reviewer on one codebase, each on its own branch. They share contracts over the bus and flag conflicts before they collide.
A long-running research swarm
A planner breaks a question into parts, several researchers take independent slices, and a synthesizer consolidates the findings into one report while you do something else.
Open source posture
ClaudeLink is released under the MIT license. The full source is on GitHub, and it ships as an npm package you install in one command. There is no account, no telemetry, and no cloud component RBJ Global operates. You can read the code, fork it, and run a build that fits your own compliance posture.
Part of the RBJ Global family
ClaudeLink shares the posture that runs across the whole portfolio: local-first, bring-your-own-keys, no telemetry, and a strong preference for keeping data and reasoning on the user's side of the wire. It is the open-source, developer-facing expression of that idea, applied to teams of AI agents.
Free, open source, no account.
Install with one command, restart your terminals, and the tools appear. MIT licensed. Everything runs on your own machine.
See the full product site, the documentation, and the quickstart:
Visit claudelink.ai →Source on github.com/RBJGlobal/claudelink.
More from RBJ Global
Clawdemy
Free AI education, narrated for life on the move.
Read more·clawdemy.org
Clawless Computer
A calm, capable desktop home for AI work.
Read more·clawless.ai
WhisprDesk
Local-first transcription. Your voice never leaves your Mac.
Read more·whisprdesk.com
Trading Agents Lab
An open-source lab for studying how multi-agent LLM systems reason about markets.
Read more·tradingagentslab.ai
iLoveMD
Free Markdown viewer and converter. Runs 100% in your browser.
Read more·iluvmd.com