@monora-ai/mcp

Official read-only MCP server for Monora - the self-hosted "company brain".

It lets an AI agent (Claude Code, Cursor, Codex, ...) search and read only the folders your Monora access key authorizes, served over HTTPS from your own Monora instance. Nothing is copied to your machine, and it cannot write.

claude mcp add monora \
  --env MONORA_URL=https://git.your-company.example \
  --env MONORA_TOKEN=<your-access-key> \
  -- npx @monora-ai/mcp

What it exposes

A small set of read-only tools to the agent: search, read_file, list_folders, list_files. Each call is proxied to your Monora instance, which returns only what your key authorizes.

What it does NOT do

• It talks only to the instance in MONORA_URL (your own Monora server). No third-party hosts, no telemetry.
• It is read-only: no writing, no deleting, no pushing. To edit, use @monora-ai/connector instead, which gives you a local editable tree.
• It only ever returns folders your key is authorized for. Access is enforced server-side.
• It stores nothing on disk.

Security model

• MONORA_TOKEN is a bearer access key, scoped per user and revocable at any time from your Monora instance's settings. Keep it out of shared logs and chats; if it leaks, revoke and reissue.
• Your Monora instance is self-hosted - your data stays on infrastructure you control.

Requirements

Node >= 20.

Built and maintained by the Monora team. Questions or issues: https://monora.ai.