The MongoDB MCP Server allows you to interact with MongoDB clusters using natural language queries from AI clients that support MCP. This page describes the MCP Server tools.
Overview
The MongoDB MCP server provides the following tools categories:
Atlas tools, which perform operations on Atlas system resources, like organizations, projects, clusters, database user accounts, and retrieving performance recommendations.
Local Atlas tools, which allow you to list, connect to, create, and delete local Atlas deployments.
Database tools, which perform operations such as inserting, updating, and deleting documents, and running queries and aggregation pipelines.
MCP Server Atlas Tools
The Atlas tools are only available if you have set up Atlas API credentials as shown in MongoDB MCP Server Configuration.
The following table describes the Atlas tools:
MCP Server Atlas Tool Name | Description |
|---|---|
| Returns a list of Atlas organizations. |
| Returns a list of Atlas projects. |
| Creates a new Atlas project. |
| Returns list of Atlas clusters. |
| Returns information about a specific Atlas cluster. |
| Creates a free Atlas cluster. |
| Connects to an Atlas cluster using the configured service account. If you configured the MCP server without specifying a connection string, this tool creates a temporary database user with a random password to establish the connection. For details, see Tool Details. |
| Returns information about the IP and CIDR ranges that can access an Atlas cluster. |
| Configures the IP and CIDR access list for an Atlas cluster. |
| Returns a list of Atlas database users. |
| Creates an Atlas database user. |
| Returns a list of alerts for an Atlas project. |
| Returns Performance Advisor recommendations
for an Atlas cluster. Supports operations for suggested indexes, drop index suggestions,
slow query logs, and schema suggestions. Requires To learn more, see Performance Advisor Tool. |
MCP Server Local Atlas Tools
You can use the MCP Server with local Atlas deployments. To use the MCP Server tools with local Atlas deployments, you must install Docker. For an introduction to local Atlas deployments, see Create a Local Atlas Deployment.
The following table describes the local Atlas tools:
MCP Server Local Atlas Tool Name | Description |
|---|---|
| Lists local Atlas deployments. |
| Creates a local Atlas deployment. To run this tool, you must disable read only mode. |
| Connects to a local Atlas deployment. |
| Deletes a local Atlas deployment. To run this tool, you must disable read only mode. |
For examples that run the local Atlas tools, see Local Atlas Deployments.
MCP Server Database Tools
The following table describes the database tools:
MCP Server Database Tool Name | Description |
|---|---|
| Connects to a MongoDB cluster. |
| Runs a MongoDB database query. |
| Runs a MongoDB aggregation pipeline. |
| Returns the number of documents in a collection. |
| Returns statistics describing the execution of the winning plan chosen by the query optimizer for the evaluated method. |
| Adds documents to a collection. If you specify a Voyage AI API key in your MCP configuration, the server can automatically generate vector embeddings from text and include them in the inserted documents. |
| Creates an index on a collection. This tool supports creating vector search indexes. |
| Removes a vector search index from a collection. |
| Modifies a single document in a collection. |
| Modifies multiple documents in a collection. |
| Changes the name of a collection. |
| Creates a new collection. |
| Removes documents from a collection. |
| Deletes a collection from a database. |
| Deletes a database. |
| Drop an index for the provided database and collection. |
| Returns a list of all databases available through the current connection. |
| Returns a list of collections in a database. |
| Returns information about collection indexes, including vector search indexes. |
| Returns collection schema information. |
| Returns collection size in megabytes. |
| Returns database statistics. |
| Saves the results of a query or aggregation pipeline in JSON
format to a file on the computer that runs the MCP Server.
The results are also accessible through the |
| Returns the most recent logged |
| Switch to a different MongoDB connection. |
Tool Details
For additional information about specific MCP tools, see the following sections.
Vector Search Support
Vector search support in MCP is available as a preview feature.
To enable this feature, set the previewFeatures flag or
MDB_MCP_PREVIEW_FEATURES environment variable to
search in your MCP configuration.
To learn more, see MongoDB MCP Server Configuration Options.
The MongoDB MCP Server supports MongoDB Vector Search. You can create and manage vector search indexes, generate embeddings, and run semantic search queries through natural language prompts. The following table summarizes key features.
Use Case | Example Prompts | Relevant Tools |
|---|---|---|
Manage indexes | Create a vector search index on the sample_db database and products collectionShow me all vector search indexes on the products collectionDrop the vector search index named vector_index | create-indexcollection-indexesdrop-index |
Insert documents with automatic embeddings | Insert these documents into the products collection and embed their descriptions | insert-many |
Vector search queries | Search for documents semantically similar to this descriptionFind me related products filtered by price range | aggregate |
Use the following resources to learn more:
For detailed usage examples and sample outputs, see Vector Search.
To configure the MCP Server for vector search, see Vector Search Options.
To learn more about vector search, see MongoDB Vector Search Overview.
Index Management
The following tools allow you to manage vector search indexes:
collection-indexes: Lists all indexes on a collection, including vector search indexes, and provides index status information.create-index: Creates a new vector search index on a collection.drop-index: Deletes a vector search index from a collection.
Note
To update a vector search index, drop the existing index and create a new one.
Automatic Embedding Generation
If you configure the MCP Server with a Voyage AI API key, the server can automatically generate embeddings in the following ways:
Generate embeddings for documents: Embeds text fields in documents when using the
insert-manytool.Generates embeddings for queries: Embeds the search query when running vector search queries with the
aggregatetool. Specifically, the server generates embeddings for thequeryVectorparameter in$vectorSearchaggregation queries.
The MCP Server supports the following Voyage AI embedding models:
voyage-3-largevoyage-3.5voyage-3.5-litevoyage-code-3
To learn more about Voyage AI models, see the Voyage AI documentation.
Note
By default, the MongoDB MCP Server validates that fields with vector search indexes
contain valid vector embeddings to prevent breaking vector search indexes.
To disable this behavior, set the disableEmbeddingsValidation option to true.
To learn more, see Vector Search Options.
Considerations
The MongoDB MCP Server supports pre-filtering vector search queries. To learn more, see MongoDB Vector Search Pre-Filtering.
The MCP server does not support the quantization field for
vector search indexes.
Performance Advisor Tool
The atlas-get-performance-advisor tool allows you to access
Performance Advisor recommendations
through natural language queries. This tool helps you identify performance
optimization opportunities by analyzing slow queries and suggesting improvements.
When performing slow query analysis, the MongoDB MCP Server retrieves a sample of slow queries, capped at 50 queries. The sample includes up to 50 most recent slow queries that match any specified conditions in your prompt to ensure optimal performance and response times.
Note
This tool requires Project Read Only access or higher and an M10+ cluster.
It is available with the --readonly flag.
Use Case | Example Prompts | Performance Advisor Operation |
|---|---|---|
Analyze Slow Queries | Show me my slow queriesWhat is slowing down my cluster?Show me queries that are longer than 5 secondsShow me slow writes in the website.users namespace | |
Index Suggestions | Are there any indexes I should create to improve performance?What indexes do you recommend I drop? | |
Schema Advice | Show schema recommendations for my clusterHelp me optimize my database schema |
For detailed usage examples and sample outputs, see Performance Optimization.
Connecting without Atlas Connection String
If you configure the MCP server without specifying a connection string
to an Atlas cluster, the atlas-connect-cluster tool
creates a temporary database user
to establish connection to the cluster
by using the configured Atlas API service account credentials.
The temporary database user has the following characteristics:
Randomly generated username and password.
Automatically expires after 12 hours.
Role assigned based on how you configured the MCP Server:
readAnyDatabaseif you enabled read-only mode or disabled thecreate,delete, andupdatetool categories.readWriteAnyDatabaseif the server has full permissions.
Note
The MongoDB MCP Server stores user credentials in memory only and never returns or exposes the credentials in the LLM context.
Learn More
To disable specific tools and restrict the MCP Server to read-only mode, see MongoDB MCP Server Configuration.
To see some MCP Server example natural language prompts, see MongoDB MCP Server Usage Examples.