mongodb-mcp-setup

Official MongoDB schema-design rules covering 3 anti-patterns (unnecessary collections, excessive $lookup, unused indexes), 4 fundamentals (embed vs reference, document model, $jsonSchema validation, 16MB limit), and 11 design patterns (bucket, computed, extended reference, polymorphic, schema versioning, time-series).

MongoDB skill that diagnoses slow queries and recommends indexes using the MongoDB MCP server (collection-indexes, explain, find) and Atlas Performance Advisor. Prefers indexing over rewriting, favors fully-covered indexes, and only activates when the user explicitly asks for performance help.

MongoDB client-connection optimization skill — picks pool size, timeouts, and idle/monitoring settings per workload (serverless, OLTP, OLAP) across all official drivers (Node.js, Python, Java, Go, C#, Ruby, PHP). Always gathers environmental context first and justifies each parameter, instead of applying arbitrary values.

Generates read-only MongoDB find queries and aggregation pipelines from natural language, using MCP tools to gather indexes, collection schema, and sample documents first, then emitting stringified Extended-JSON queries that respect index coverage and avoid `$where`, `$regex` on unindexed fields, or redundant operators. Excludes Atlas Search / Vector Search (see mongodb-search-and-ai).

Guides MongoDB users through Atlas Search (lexical/full-text), Vector Search (semantic/RAG), and Hybrid Search (`$rankFusion` MongoDB 8.0+, `$scoreFusion` 8.2+) via the MongoDB MCP server — inspecting indexes/schema/cluster version, choosing the right search type, explaining the proposed index before creating it, and avoiding anti-patterns like `$regex`/`$text` for search workloads.

Official MongoDB agent skills for agentic workflows, providing MCP server integration and connection configuration for Atlas. Compatible with Claude Code, Cursor, and Gemini CLI.