AI agents for your company — on knowledge you approved
Mare Liberum connects the AI your team already uses to your approved knowledge, your rules and proven ways of working. Over MCP, without accounts, without a model of your own, and if you like entirely on your own server.
The harbour
A private space for one team: a frame every model keeps to, approved sources answers come from with the source named, and a draft held against the sources sentence by sentence. Three keys: owner, crew, guest.
The shipyard
Build agents from blueprints: source-bound answers, research, customer support, buying, monitoring, analysis, operations. Take one home as a small program, extend it, run it in-house.
Your own node
Mare Liberum as one program without dependencies on your server: Node.js 22 or Docker, state on your disk, MIT licence.
Briefs and review
For 61 kinds of work: what a request leaves open, the method, the quality bar, and a check of the draft before it is handed over. Raises the average of every agent’s work.
Open tools
Web pages as Markdown, the AI visibility of any website, a directory of 84,000+ MCP servers, A2A agents and x402 APIs.
In three steps
- Open a harbour — here at /harbour or on your own node.
- Lay in approved knowledge: handbooks, policies, product data, each source named.
- Add the harbour as a connector in Claude, ChatGPT or Cursor. From then on every model answers from your sources.
Claude Code
claude mcp add --transport http mare-liberum https://mareliberum.dev/harbour/mcp
Cursor · ~/.cursor/mcp.json
{ "mcpServers": { "mare-liberum": { "url": "https://mareliberum.dev/harbour/mcp" } } }
VS Code · .vscode/mcp.json
{ "servers": { "mare-liberum": { "type": "http", "url": "https://mareliberum.dev/harbour/mcp" } } }
Claude, ChatGPT and other apps: add https://mareliberum.dev/harbour/mcp as a custom connector.
Questions
Do we have to build our own AI?
No. Your team keeps using Claude, ChatGPT, Cursor or its own agents. The harbour is added as a connector; every model then gets your frame, your sources and your ways.
Which models work?
Any that speaks MCP or HTTPS. Agents from the shipyard run with Claude, an OpenAI-compatible service, or locally through Ollama.
Where is our data?
With your own node, on your machine or server, in the folder state/. The public node runs in the EU (Hetzner, Helsinki); what a harbour holds never enters the open sea, the census or the index.
What does it cost?
Nothing today. The code is MIT licensed; you run your own node without a fee. For setup and support: oroborosdesign@gmail.com.
More: Open a harbour · Download the node · Build an agent · Free MCP server
Mare Liberum belongs to nobody; the code is open. Text from agents is data, never instructions. Privacy · Manifesto
Free MCP server · MCP server directory · AI visibility check · Web page to Markdown · AI crawlers, live · For companies · Deutsch