MongoDB Atlas MCP Server is an MCP server that provides tools for managing MongoDB Atlas cloud infrastructure through AI clients. It connects directly to the MongoDB Atlas administration API, allowing developers, DevOps engineers, and database administrators to manage Atlas projects without leaving their code editors or chat assistants. Users can provision new MongoDB clusters in existing projects, list projects and clusters, retrieve database connection strings, and add administrative users with the atlasAdmin role. The server also provides capabilities to adjust project network access rules, making it possible to configure IP whitelisting alongside resource creation. By exposing these management capabilities via the Model Context Protocol, the server enables language models to orchestrate Atlas cloud operations, verify infrastructure state, and automate administrative provisioning workflows directly through standard MCP tooling.
Category: Cloud & Infrastructure
Tags: atlas, cloud-database, database-management, mongodb, nosql
Visit MongoDB Atlas MCP Server
claude_desktop_config.json, Cline's settings, or ~/.cursor/mcp.json), add the server entry. 3. Configure the server definition with command npx, args ["mcp-mongodb-atlas"], and environment variables ATLAS_PUBLIC_KEY and ATLAS_PRIVATE_KEY set to your API keys. 4. Restart your MCP client to load the tools.Part of MCP Servers
You can install it using npx by running npx mcp-mongodb-atlas, or install it locally via npm using npm install mcp-mongodb-atlas. Add the command and your Atlas API keys to your MCP client configuration.
The server allows AI clients to manage MongoDB Atlas infrastructure. It includes tools to list projects and clusters, create new clusters, retrieve cluster connection strings, create database users with admin roles, and configure project network access settings.
The server works with standard MCP clients including Claude Desktop, Cursor, and the Cline extension for VSCode. It communicates over standard input/output (stdio) and can also run directly from the command line.
Yes, it is open source and licensed under the MIT license. You can view the source code and build it locally from its public GitHub repository.