mongodb-js/mongodb-mcp-server

A Model Context Protocol server to connect to MongoDB databases and MongoDB Atlas Clusters.

MongoDB MCP Server – Quick Overview

What it is – A Model Context Protocol (MCP) server that lets AI agents talk to MongoDB databases (both self‑hosted and MongoDB Atlas). It implements a standard protocol so tools like Claude, Cursor, GitHub Copilot, Grok, etc., can issue database commands (find, aggregate, create‑index, etc.) without the agent having to embed raw connection logic.

Why it matters – Enables “data‑aware” LLM agents: the model can ask the server for data, run queries, or even create resources in Atlas, all through a safe, read‑only‑by‑default interface.

Key Features (as described in the README)

Feature Details
Supported AI client integrations Official plugins for Cursor, VS Code, Claude, Codex, GitHub Copilot CLI, Grok, Gemini. Two plugins are shipped: mongodb-atlas (remote Atlas MCP via OAuth) and mongodb (local server for any MongoDB deployment).
Command‑line / Docker deployment Run locally with npx -y mongodb-mcp-server@latest --readOnly or via Docker (mongodb/mongodb-mcp-server:latest). Environment variables supply either a MongoDB connection string or Atlas Service Account credentials.
Read‑only default The server starts with --readOnly to protect data; remove the flag to enable writes.
Rich toolset Exposes many MongoDB operations as MCP tools (e.g., find, aggregate, create-index, atlas-create-cluster, atlas-get-performance-advisor, etc.).
Security recommendations Prefer environment variables for secrets, grant minimal Atlas API permissions, and keep the server in read‑only mode unless writes are explicitly needed.
Configuration flexibility Supports JSON‑style MCP configuration files, command‑line args, env vars, and a Docker‑based launch.
HTTP compatibility A separate library guide describes how custom HTTP clients can talk to the server (headers, redirects, payload limits).
Cloud deployment Mention of Azure deployment (details in the docs).

Typical Workflow

  1. Install the appropriate plugin for your AI assistant (e.g., mongodb-atlas in Claude Marketplace).
  2. Provide credentials – either a MongoDB URI (MDB_MCP_CONNECTION_STRING) or Atlas Service Account client ID/secret (MDB_MCP_API_CLIENT_ID / MDB_MCP_API_CLIENT_SECRET).
  3. Start the server – via npx or Docker. The server runs as a subprocess that the AI client talks to using the MCP protocol.
  4. Issue commands – The agent can call tools like find, aggregate-db, atlas-create-cluster, etc., and receive structured JSON results.
  5. Optional write access – Remove --readOnly if the agent needs to modify data, after confirming security policies.

Who Might Use This?

  • Developers building AI‑augmented applications that need to fetch or manipulate data stored in MongoDB.
  • Data‑science teams wanting LLMs to explore their MongoDB datasets safely.
  • Ops teams that let AI agents provision or inspect Atlas clusters via the atlas‑* tools.

Getting Started (quick commands)

# Quick local start (read‑only)
npx -y mongodb-mcp-server@latest setup   # interactive wizard
# Or run directly
npx -y mongodb-mcp-server@latest --readOnly

# Docker alternative
docker run --rm -i -e MDB_MCP_CONNECTION_STRING='mongodb://localhost:27017/myDb' \
    -e MDB_MCP_READ_ONLY='true' mongodb/mongodb-mcp-server:latest

Where to Learn More

  • Full documentation: https://www.mongodb.com/docs/mcp-server/
  • Plugin marketplaces: Cursor, Claude, GitHub Copilot, Grok, Gemini.
  • Library guide for custom HTTP clients: MCP_SERVER_LIBRARY.md.

All information above is taken directly from the repository’s README; no additional features have been inferred.

Related

  • Project
    motherduckdb/mcp-server-motherduck

    A self‑hosted MCP server that wraps DuckDB (local, S3, or MotherDuck) so AI assistants can run SQL queries via the Model‑Chat‑Protocol. Supports stdio or HTTP, read/write control, Docker deployment, and integrates with Claude, Gemini, Cursor, etc.

  • Project
    SonarSource/sonarqube-mcp-server

    The SonarQube MCP Server is a Docker‑based bridge that lets AI coding assistants (Claude, Gemini, Copilot, Cursor, VS Code, etc.) query a SonarQube instance for code‑quality and security findings. Set it up via the online configuration generator or by running the `sonarsource/sonarqube-mcp` image, provide your SonarQube token (and optional org or server URL) as environment variables, and add the generated JSON snippet to the agent’s MCP configuration. The server then answers the agent’s requests for project‑wide or snippet‑level analysis, enabling trustworthy, up‑to‑date static‑analysis data inside AI‑augmented development workflows.

  • Project
    googleapis/mcp-toolbox

    MCP Toolbox is an open‑source server that lets AI agents and IDEs talk to enterprise databases via the Model Context Protocol. It offers ready‑made database tools and a framework for defining custom, secure SQL/semantic‑search tools, with SDKs for Python, JS/TS, Go, and Java.

  • Project
    designcomputer/mysql_mcp_server

    A Model Context Protocol (MCP) server that enables AI applications to securely explore and query MySQL databases through a structured interface.

  • Project
    a-bonus/google-docs-mcp

    A Node.js MCP server that lets AI agents (Claude Desktop, Cursor, etc.) read, edit, and manage Google Docs, Sheets, Drive files, Gmail messages, Calendar events, and Apps Script projects via natural‑language‑style tool calls. It works locally via `npx` after OAuth setup or can be deployed to Cloud Run for team‑wide use.