Skip to main content
MCP server CX Codex

Lumberroom — Claude Code MCP server: features and setup

Lumberroom is an MCP server for Claude Code that gives agents shared facts to search, save, and correct. Per-client policies control which facts each tool can access.

Apache-2.0 ★ 1 updated 2026-11-15

Install into Claude Code · Platform CLI: remote server All methods and platforms ↓
claude mcp add --transport http lumberroom https://mcp.lumberroom.cloud/mcp
Open on GitHub

What Lumberroom does

Lumberroom lets AI tools read and update a shared store of facts. A structured registry keeps canonical keys and provenance, while access policies decide which records each client can see.

You can host the server on your own infrastructure; it uses PostgreSQL 16 with pgvector. Claude Code connects over HTTP with the listed MCP command. In a self-hosted token setup, the server binds to 127.0.0.1:8787; losing secrets/lumberroom-kek makes private records unreadable.

What it helps with

  • find facts saved in earlier agent sessions
  • share a stored fact across connected clients
  • correct a record while retaining its provenance
  • control which facts each client can access

Example prompts for the agent

  • Find the Ubuntu version I saved for my desktop.
  • Save that the test server now uses PostgreSQL.
  • Correct the Node.js version in shared memory.

How to install Lumberroom

Platform CLI: remote server

claude mcp add --transport http lumberroom https://mcp.lumberroom.cloud/mcp

Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more

Questions about Lumberroom

How do I install Lumberroom in Claude Code?

Run claude mcp add --transport http lumberroom https://mcp.lumberroom.cloud/mcp to add the HTTP MCP server.

What does Lumberroom do?

It gives AI agents a shared store where they can search, save, and correct facts. Per-client policies control access to those records.

What does a self-hosted Lumberroom server need?

The server uses PostgreSQL 16 with pgvector. Client access requires credentials, either a static token or OAuth, depending on the connection method.

Similar MCP servers

MCP servers catalog →

Project sponsors

Companies that keep this analytics open