FireMCP registers 5 tools with basic schemas and descriptions. All tools have input schemas with type definitions and descriptions present. However, descriptions are minimal (10-50 chars), falling below the 50-200 char baseline for LLM-optimized descriptions. Output schemas are referenced but not visible in source. Parameter descriptions exist but lack detail on constraints, formats, and error recovery. No tool annotations (readOnlyHint, destructiveHint) despite clear risk levels. Error handling and validation logic not visible in provided code. Tool composition is sound (each does one thing), but descriptions need expansion to guide LLM selection and usage.

• Expand all tool descriptions to 50-200 chars. Example: 'get_document' → 'Retrieve a Firestore document by path (e.g., users/userId). Returns document data and metadata. Use this to fetch a single record before updating or deleting it.'
• Add tool annotations to server.registerTool() calls: readOnlyHint=true for get_document/query_collection, destructiveHint=true for delete_document, idempotentHint=true for set_document (with merge=false).
• Document filter operators in query_collection description: 'operators: ==, <, <=, >, >=, !=, array-contains, in, array-contains-any'.
• Verify and document output schemas: ensure get_document returns {data, exists, path}, set_document returns {path, success}, query_collection returns {documents: [{id, data}], count, hasMore}.
• Add error recovery guidance: 'If document not found, check path format (collection/docId). If permission denied, verify Firestore Security Rules allow this operation.'