The MongoDB MCP server continues to evolve rapidly as the MCP ecosystem matures and developer needs accelerate. We're keeping pace in this fast-moving space, continuously delivering new capabilities that expand what's possible. This momentum reflects both the speed of change of AI tooling and our commitment to meeting developers where they are. Below are some of the most impactful features from recent months, with details on what each enables.

Performance advisor tools

The MCP server now directly integrates with the Performance Advisor tools found within MongoDB Atlas. Instead of needing to use the Atlas CLI or Atlas UI, you can get that information right in your development environment from the MCP server.

This new capability is enabled specifically by four new functions:

• listClusterSuggestedIndexes() analyzes your queries and recommends indexes that would improve performance. It's based on actual usage patterns, so the suggestions are relevant to what you're building.

• listDropIndexes() identifies indexes that aren't being used. Unused indexes take up space and slow down writes without helping query performance, so knowing which ones to remove is useful.

• listSchemaAdvice() looks at your data structure and suggests improvements. Schema design impacts everything from query speed to storage efficiency, and having recommendations based on real data helps.

• listSlowQueries() surfaces queries that perform poorly so you can identify bottlenecks. This tool helps make finding slow queries straightforward and easy, instead of digging through logs.

These functions let you work on performance optimization without context switching between different tools and interfaces. They help improve the performance of your database with the MCP.

Try it out:

Create vector search indexes

The CreateIndexTool class now supports vector search indexes alongside regular MongoDB indexes. You use the same tool for both types, which simplifies the interface.

The implementation determines what type of index to create based on the parameters you pass in. Vector search capabilities are currently behind a feature flag during the initial release phase. The flag will be removed once the feature matures.

Having one tool handle multiple index types means you don't need to learn separate commands or remember different syntax depending on whether you're creating a traditional index or a vector search index.

Atlas local cluster creation

The MCP server can now create and manage local MongoDB clusters. This removes several manual steps from local development setup. You can create a local cluster using the mongodb-atlas-local image without setting anything up outside your workflow. The MCP server handles cluster creation, connection, and management.

You can list and inspect local clusters to see what's running, and delete them when you're finished. This keeps your development environment organized without manual cleanup. This capability means you don't need to install MongoDB separately, configure it, and then provide connection details to your tools. The MCP server handles that process.

Try it out and check out more Atlas Local deployment examples.

Automatic embeddings for insert-many

The insert-many tool now supports automatic embedding generation for fields with vector search indexes. When you insert documents, the function can now generate embeddings using the best embedding models made by Voyage AI.

This removes the need to manually generate embeddings before insertion. The tool checks your vector search configuration and generates embeddings for the appropriate fields during the insert operation. To enable automatic embedding generation, simply provide the optional parameters to the function when needed. Try it out now!

Assistant tools

MongoDB is adding assistant tools that integrate with the MongoDB assistant for knowledge search and retrieval. These tools will allow the MCP to utilize the MongoDB knowledge base more efficiently than ever.

• list-knowledge-sources will retrieve available data sources from the assistant, which helps with filtering searches to relevant information.

• search-knowledge will run natural language queries against the MongoDB assistant knowledge base and return relevant content chunks.

Wrapping up

Recent updates to the MongoDB MCP server focus on practical improvements: performance optimization tools that work within your development environment, unified index creation for both traditional and vector indexes, local cluster management that simplifies setup, and automatic embedding generation that reduces repetitive tasks. These changes address real friction points in the development process and make it easier to work with MongoDB through AI-assisted tools. Stay tuned as we continue to enhance the MCP server and make it even more useful.