Single tool with critical gaps. The tool 'find-firebase-doc' has a basic description but lacks essential parameter validation, output schema documentation, error handling guidance, and comprehensive parameter descriptions. Parameter 'request' is described generically ('used as the goal the user is trying to accomplish') without constraints, format guidance, or examples. No output schema is documented, callers cannot know what fields to expect from the QueryResult array. Error handling is entirely absent, what happens when no results are found? When the embedding service fails? When the database connection fails? The tool lacks idempotence hints, rate-limit guidance, and pagination despite querying a vector database that could return many results. Tool name 'find-firebase-doc' is adequate but could be more specific (e.g., 'search_firebase_docs' would better signal the search/lookup pattern). This server does not meet production quality baselines.

• Document the output schema explicitly: 'Returns an array of objects with fields: source (string, URL or file path to the Firebase doc), text (string, extracted content snippet). Maximum 10 results returned.'
• Enhance parameter description: 'request: The user's goal or question about Firebase (1-500 characters). E.g., "How do I set up Data Connect?", "What is the billing model?". Will be converted to embeddings for semantic search.'
• Add error handling and recovery guidance in tool description or as explicit error responses: 'If no results found: Try a more general query. If service is slow: Retry after 2 seconds. If database connection fails: The service is temporarily unavailable.'
• Document the result limit and pagination behavior: 'Returns up to 10 most relevant documents. If you need more results, refine your query for better relevance rather than requesting additional pages.'
• Add idempotence note: 'This tool is idempotent, repeated calls with the same request always return results in the same order, making it safe to retry.'
• Constrain the input parameter: use z.string().min(1).max(500).describe('...') to prevent empty strings and excessively long inputs.
• Consider renaming to 'search_firebase_docs' (more standard verb) for clearer intent signaling to LLMs.
• Add dependency hints if applicable: 'If searching for a specific Firebase feature, mention the service name (e.g., Realtime Database, Firestore) for better results.'