MongoDB Lens

Executes MongoDB queries (find, insert, update, delete, aggregate) through the Model Context Protocol, translating natural language or structured requests from Claude/LLMs into native MongoDB driver calls. Implements MCP resource and tool handlers that map incoming requests to pymongo or native MongoDB driver operations, managing connection pooling and query result serialization back to the LLM context.

Automatically discovers and exposes MongoDB database schema information (collections, indexes, field types, validation rules) as MCP resources, allowing LLMs to understand database structure without manual documentation. Queries MongoDB system catalogs (system.indexes, schema validation metadata) and constructs a queryable schema representation that Claude can reference when formulating queries.

Implements MongoDB change streams as MCP resources, allowing Claude to monitor database changes in real-time and react to insert, update, delete, and replace operations. Handles change stream lifecycle (open, filter, close) and provides event notifications that Claude can use to trigger downstream actions or maintain synchronized state.

Provides MCP tools for building and executing MongoDB aggregation pipelines, translating high-level analytical requests into multi-stage pipeline definitions. Handles stage composition ($match, $group, $project, $sort, $limit), result streaming, and error handling for complex data transformations that go beyond simple CRUD operations.

Manages MongoDB connection lifecycle through MCP, maintaining a persistent connection pool that persists across multiple LLM requests within a single conversation. Implements session reuse, automatic reconnection on failure, and proper resource cleanup to avoid connection exhaustion when Claude makes multiple sequential database calls.

Supports bulk insert, update, and delete operations through MCP, allowing Claude to perform multiple database modifications in a single atomic or ordered batch. Implements bulk write API wrappers that translate batch operation requests into MongoDB bulk write commands, with error handling for partial failures and detailed operation counts.

Provides MCP tools for creating, listing, and deleting MongoDB indexes, and allows Claude to apply query hints to optimize execution plans. Exposes index creation with configurable options (unique, sparse, TTL) and enables query hints that instruct MongoDB to use specific indexes, helping Claude learn which indexes improve query performance.

Leverages MongoDB's schema validation feature to enforce document structure constraints, exposing validation rules as MCP resources and allowing Claude to understand what documents are valid before insertion. Reads and applies JSON Schema validation rules, providing feedback when Claude attempts to insert documents that violate schema constraints.

Exposes MongoDB multi-document transactions through MCP, enabling Claude to execute sequences of operations (read, write, delete) that are guaranteed to be atomic and isolated. Implements transaction lifecycle management (begin, commit, abort) with proper error handling for transaction conflicts and automatic rollback on failure.

Exposes MongoDB text search capabilities through MCP, allowing Claude to perform full-text searches across indexed fields using MongoDB's text search syntax. Supports text index creation with language-specific stemming and stop words, and enables natural language search queries that MongoDB translates into efficient index lookups.

Exposes MongoDB geospatial query capabilities through MCP, enabling Claude to perform location-based searches using 2dsphere or 2d indexes. Supports queries like finding documents within a radius, along a route, or within a polygon, with automatic coordinate validation and distance calculation.