Server has 6 tools with basic descriptions but critical gaps in schema completeness, parameter documentation, and error handling. All tools lack input parameter descriptions (e.g., 'index' param in get_mapping has no description of format/constraints). Output schemas are undocumented, responses are raw TextContent strings with no structured field definitions. Error handling is minimal: catch-all exceptions return plain error text without recovery guidance. No tool annotations (readOnlyHint, etc.). Naming is acceptable (verb_noun pattern) but descriptions are generic and under 100 chars, missing context on when/why to use each tool. No pagination support despite search_documents accepting arbitrary DSL queries that could return thousands of results.

• Add descriptions to all input parameters. For 'index' param: 'Name of the Elasticsearch index (e.g., logs-2024-01, my-data). Use list_indices to discover available indices.'
• Document output schemas with structured field definitions. Instead of returning str(response), parse and return a dict with typed fields: {indices: [{name: str, docs: int, size_bytes: int}], total: int}
• Add pagination to search_documents: accept 'limit' (1-100, default 20) and 'offset' (default 0) parameters. Return {hits: [...], total: int, limit: int, offset: int} to enable multi-page result traversal.
• Implement actionable error messages. Catch specific exceptions (IndexNotFound, ConnectionError, etc.) and return: 'Index "my-index" not found. Available indices: [list]. Call list_indices() to discover indices.'
• Add tool annotations via fastmcp: @mcp.tool(..., readOnlyHint=True) for all read-only tools to signal to LLMs that these are safe to call without confirmation.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry