mongodb-natural-language-querying

Describe the documents you want in plain language, including the database and collection name. The skill fetches the collection schema, indexes, and sample documents through the MongoDB MCP server, then returns a validated find query or aggregation pipeline in JSON format.

Use a find query for simple filtering, sorting, projection, and limiting on one collection. Use an aggregation pipeline when you need grouping with functions like $sum or $avg, joins via $lookup, array unwinding, or multiple transformation stages.

No, it does not handle $search, $vectorSearch, fuzzy matching, autocomplete, or relevance scoring. Those require a dedicated search skill, and it also does not analyze or optimize existing queries or pipelines containing write operations.

Yes, the skill fetches collection indexes before generating the query and checks whether $match and $sort stages can use them. If no supporting index exists, it notes this in the response so you can consider creating one.

MongoDB does not error on nonexistent field names, so typos silently return empty results. This skill prevents that by validating every field name and type against the actual collection schema before generating the query.