MongoDB

MongoDB’s official MCP Server that connects AI assistants like GitHub Copilot, Claude, and others directly to your MongoDB databases and Atlas deployments.

It turns natural language prompts or automated workflows into safe, structured database operations without requiring manual query writing or deep MongoDB expertise.

Features

• 🗃️ Dual Access Modes: Connect via MongoDB connection string or MongoDB Atlas API credentials
• 🔒 Built-in Safety Controls: Read-only mode, index enforcement, and configurable tool restrictions
• 🧰 Rich Toolset: 20+ database operations (find, aggregate, insert, drop, etc.) plus 10+ Atlas management tools
• 📤 Export Support: Save query results as EJSON files accessible via URI resources
• 📊 Observability: Built-in logging to disk, stderr, or MCP client with debug resources for connectivity issues
• 🌐 Transport Flexibility: Supports stdio (default) and HTTP (with security caveats)
• 🛡️ Secure Credential Handling: Encourages environment variables over command-line args for secrets
• ⚙️ Fine-Grained Permissions: Disable tools by name, category (e.g., atlas, delete), or operation type

How to Use It

1. Add a JSON object to your MCP client’s configuration file.

Connection String Method (Recommended for Simplicity)

This is the easiest way to connect to an existing database, whether it’s running locally or on Atlas. Pass the connection string securely using an environment variable in your client’s config file.

{
  "mcpServers": {
    "MongoDB": {
      "command": "npx",
      "args": ["-y", "mongodb-mcp-server@latest", "--readOnly"],
      "env": {
        "MDB_MCP_CONNECTION_STRING": "mongodb+srv://user:[email protected]/myDatabase"
      }
    }
  }
}

Atlas API Credentials Method (For Full Atlas Management)

If you want to manage your Atlas environment (create projects, clusters, etc.), you’ll need to use API credentials from a Service Account.

First, create a Service Account in your Atlas organization’s Access Manager. Grant it the necessary permissions. Atlas will provide a Client ID and a Client Secret. Save the secret immediately, as it won’t be shown again.

Then, configure your client like this:

{
  "mcpServers": {
    "MongoDB": {
      "command": "npx",
      "args": ["-y", "mongodb-mcp-server@latest", "--readOnly"],
      "env": {
        "MDB_MCP_API_CLIENT_ID": "your-atlas-service-account-client-id",
        "MDB_MCP_API_CLIENT_SECRET": "your-atlas-service-account-client-secret"
      }
    }
  }
}

Using Docker

If you prefer not to install Node.js locally, you can run the server in a Docker container. The configuration is similar; you just change the command and args to execute a docker run command.

{
  "mcpServers": {
    "MongoDB": {
      "command": "docker",
      "args": [
        "run", "--rm", "-i",
        "-e", "MDB_MCP_CONNECTION_STRING",
        "-e", "MDB_MCP_READ_ONLY=true",
        "mongodb/mongodb-mcp-server:latest"
      ],
      "env": {
        "MDB_MCP_CONNECTION_STRING": "mongodb+srv://user:[email protected]/myDatabase"
      }
    }
  }
}

2. After saving the configuration file, restart your client. The MongoDB tools should now be available to your AI assistant.

MongoDB Atlas Tools

• atlas-list-orgs: Lists your MongoDB Atlas organizations.
• atlas-list-projects: Lists all projects within your MongoDB Atlas organizations.
• atlas-create-project: Creates a new project in MongoDB Atlas.
• atlas-list-clusters: Lists the database clusters in a specific MongoDB Atlas project.
• atlas-inspect-cluster: Provides detailed information about a specific MongoDB Atlas cluster.
• atlas-create-free-cluster: Deploys a new free-tier M0 cluster in MongoDB Atlas.
• atlas-connect-cluster: Connects the server to a specified MongoDB Atlas cluster.
• atlas-inspect-access-list: Shows the IP/CIDR ranges that have access to your MongoDB Atlas clusters.
• atlas-create-access-list: Configures the IP/CIDR access list for your MongoDB Atlas clusters.
• atlas-list-db-users: Lists all database users for a project in MongoDB Atlas.
• atlas-create-db-user: Creates a new database user within a MongoDB Atlas project.
• atlas-list-alerts: Lists all alerts for a given project in MongoDB Atlas.

MongoDB Database Tools

• connect: Establishes a connection to a MongoDB instance.
• find: Executes a find query on a specified collection.
• aggregate: Runs an aggregation pipeline against a collection.
• count: Returns the total number of documents in a collection.
• insert-one: Adds a single document to a collection.
• insert-many: Adds multiple documents to a collection at once.
• create-index: Creates a new index on a collection.
• update-one: Modifies a single document that matches a filter.
• update-many: Modifies all documents that match a filter.
• rename-collection: Changes the name of a collection.
• delete-one: Removes a single document from a collection.
• delete-many: Removes all documents from a collection that match a filter.
• drop-collection: Deletes an entire collection from a database.
• drop-database: Deletes an entire database.
• list-databases: Lists all available databases for the current connection.
• list-collections: Lists all collections within a specific database.
• collection-indexes: Describes the existing indexes for a collection.
• collection-schema: Analyzes a collection and describes its schema.
• collection-storage-size: Reports the storage size of a collection in megabytes.
• db-stats: Provides statistics for a specific database.
• export: Exports the results of a query or aggregation to an EJSON file.

FAQs

Q: Do I need both a connection string and Atlas API keys?
A: No, you typically use one or the other. Use the connection string for direct database queries (find, aggregate, etc.). Use the Atlas API credentials when you need to perform administrative actions on your Atlas account, like creating projects or clusters.

Q: My connection isn’t working. How can I debug it?
A: First, double-check your credentials in the configuration. Ensure your IP address is on the Atlas IP Access List if you’re connecting to an Atlas cluster. The server also provides a debug://mongodb resource that tracks the last connectivity attempt and any errors, which can be very helpful. Also, check the server logs, which can be configured to output to disk or stderr.

Q: Can I run this as a shared service for my team?
A: Yes, the server supports an HTTP transport mode (--transport http). This allows you to run it on a server and have multiple clients connect to it over the network. However, you must implement proper security measures like authentication and encryption, as the HTTP transport is not secure by default. It’s best to place it behind a firewall or within a private network.