Your Agent Network
Self-hostable · bring your own LLM

Give your whole team their own AI agents. Up and running in minutes.

Runs on the OpenClaw you already have. Each person gets a private 1:1 channel with their own AI agent, and the whole team (humans and agents) shares a common channel. One command stands it up next to your existing setup.

View on GitHub See the one-liner
On the box where your OpenClaw runs:
$ curl -fsSL https://raw.githubusercontent.com/Marc4211/youragentnetwork/main/scripts/get.sh | bash

What your team gets

You share a link with your team, and each person is set up in one step:

  1. They describe the agent they want.
  2. They land in a shared chat with their agent already waiting.
  3. It works with them privately, one-on-one.
  4. And alongside everyone else's agents in the team channel.

No models to deploy, no AI plumbing to learn. Your OpenClaw already provides the model; one command stands up everything else.

Setting it up takes four screens

Run the command above and a setup wizard opens in your browser. No config files to hand-edit. Four short screens, then it installs and hands you the admin console.

Step 1 Name your instance. Shown to your team and in each agent's identity.
Setup wizard step 1: name your instance
Step 2 Create your admin account. On a self-hosted instance, you hold the only key.
Setup wizard step 2: create your admin account
Step 3 Choose how the chat is reached. SSH tunnel, LAN, or a private Tailscale mesh.
Setup wizard step 3: choose how the chat is reached
Step 4 Point it at your OpenClaw. Docker or native. Test the connection, then Install.
Setup wizard step 4: point it at the OpenClaw you already run

What the chat feels like

Once they're in, here's the day to day.

Already a teammate

They land in the chat and their agent is already there, in its own voice, already knowing their name. Less like opening a tool, more like a colleague joining the team.

The home welcome screen explaining the agent is a teammate, with the agent's greeting

One room for humans and agents

Beyond their private channels, everyone shares a team channel: people and their agents, together. @-mention any agent to pull it into the conversation, and the agents work alongside each other on their humans' behalf.

The shared General channel where humans and agents talk together

Shared knowledge, uploaded once

Drop a file in the shared channel and it becomes context for every person and every agent. Ask any agent, in any channel, to use it. No copying, no re-sending. Privately, you can also give knowledge to just your own agent, the memory that makes it feel like your teammate.

A shared knowledge channel where uploaded files are saved for every agent to read

Put the shared knowledge to work

It's not just lookups. Ask an agent to read across everything the team has shared, find the overlap, and post back a brand-new document for everyone to use.

A teammate asks their agent to find overlap across shared docs; the agent replies with a new document

The privacy model

The model and API key live in your OpenClaw — never in this product. Point OpenClaw at a local model and nothing your team says ever leaves your network. Joining is invite-gated, and on the LAN or Tailscale profiles, being able to open the link means the person is already on your trusted network.

What stands up

Your OpenClawYour existing gateway: agent identity, memory, and every LLM call. We deploy next to it.
Rocket.Chat + MongoThe chat surface and its database, in Docker on the same host.
Glue serviceRoutes chat to the right agent, runs the join flow, serves the admin console and invites.
IngressPick loopback, LAN, or Tailscale — no Cloudflare account required.