Cartographer KB MCP

Cartographer is a shared git wiki that agents read and extend through MCP tools while the server keeps it from breaking

  • 63 ★ on the repository
  • Last commit
MCP server

Extended access

Why this level

  • The server writes to a git repository and distributes skills, hooks and instructions to agent clients, and plugins gain extended abilities after your approval
  • The demo server is public and its messages are visible to all visitors
All reasons and checks

beppetemp/cartographer

Install

Manual install

claude mcp add --transport http cartographer-demo https://cartographer-demo.beppetemp.com/mcp

A live demo with no sign-in or token.

Access and checks

How we review

Extended accessThe tool runs code or makes network calls and only reads external data.

  • The server writes to a git repository and distributes skills, hooks and instructions to agent clients, and plugins gain extended abilities after your approval
  • The demo server is public and its messages are visible to all visitors

This is third-party code. Review the repository files before installing.

What it does

Cartographer is a single Go binary that keeps a knowledge base as a git repository of Markdown and YAML files. Agents in Claude Code, Codex, OpenCode and other clients read and extend it only through MCP tools. The server checks schema, links and style, makes one commit per write and keeps a signed log, so an agent cannot leave the base broken. The same base configures the agents themselves: skills, subagents, hooks and standing instructions live in it and reach each client in its own format. There are plugins with approval, SOPS-encrypted secrets, token-based roles for teams and an Atlas web UI with a 3D graph. A live demo server can be tried without installation.

Who it is for. Teams where every colleague has their own agent and what one learns must reach the others.

Good fit when

  • Project knowledge accumulates in a shared wiki and an agent tomorrow knows what a colleague's agent found yesterday
  • Skills and instructions for Claude Code, Codex and Kiro live in one place instead of manual copying
  • The base is checked for broken links and stale pages with one command
  • Secrets are referenced from notes and resolved on demand and never stored in plain text

Not a fit when

  • You need ordinary RAG over finished documents where the agent does not need to write
  • There is no git repository to hold the base
  • One note in an instructions file is enough

Example request

What does our knowledge base know about release launches, and run a health check for broken or stale pages

MCP

Transport
stdio, http
Authentication
not required

Limitations

Git and an empty repository you control are required for the first base. Install through Homebrew `brew install beppetemp/tap/cartographer` or a `curl | sh` script, then `cartographer setup` shows a plan and only then changes the system. The server installs as a local service, a container or a Kubernetes workload. Plugins are enabled only after your approval, writes go through MCP tools, so editing Markdown files by hand around the checks is best avoided. The demo server is open to everyone and keeps visitors' messages.

How to disable. Stop and remove the service with your install tooling (`brew uninstall cartographer`) and delete the server entries from clients. The knowledge base repository stays your ordinary git repository.

FAQ

How can I try it without installing?

Connect the demo `https://cartographer-demo.beppetemp.com/mcp`, for example `claude mcp add --transport http cartographer-demo https://cartographer-demo.beppetemp.com/mcp`. No sign-in or token is needed.

How do I install it for myself?

Run `brew install beppetemp/tap/cartographer`, then `cartographer setup`: the wizard shows a plan, creates the base and connects the clients it finds.

How is it better than a Markdown folder?

Writes go only through a server that checks schema and links, makes a commit per change and keeps a signed log. An agent cannot leave the base broken.

What is Atlas?

A read-only web UI in the same binary: a 3D graph of the base, search, what agents changed and what needs fixing.

Does it suit a team?

Yes, several bases behind one server, token-based roles for bases and maps, git as the sync layer and guided conflict resolution.

Discussion

No questions or reviews yet.

Official

BrainAPI turns text into a knowledge graph on your server and serves agents over MCP, REST and a console with fact-path answers

MCP serverBroad access551Repository stars

Knowledge-graph memory for agents: Claude Code and Codex plugins, an MCP server and a Python library

PluginExtended access33.3KRepository stars
Official

A CLI and MCP server that builds a local wiki with a knowledge graph from documents, code and links, so the agent searches it instead of scanning files

CLIBroad access735Repository stars
Official

Shared memory for fleets of AI agents with access scopes, an audit trail and a knowledge graph: an MCP server, Docker and a cloud

MCP serverBroad access557Repository stars

A personal LLM wiki with a knowledge graph, cited RAG and memory, served to Claude Code and Cursor by an MCP server

MCP serverExtended access276Repository stars

An MCP server and REST API with shared memory for Claude Code, Codex and Gemini: Obsidian notes, sessions, bug fixes, lean search.

MCP serverExtended access190Repository stars

README in short

The README starts with a live demo server and a three-step agent connection. It then explains the motivation, features, quick start, installation and how teams, plugins and secrets work.

README badge

Report a problem