The server provides 5 read-only tools for querying OpenObserve. Tool names follow verb_noun convention well (search_sql, search_logs, list_streams). All tools have descriptions and input schemas are present with type definitions. However, descriptions are terse (mostly 10-15 words), parameter descriptions lack actionable constraints (ranges, formats, valid values), and output schemas are not documented. Error handling is basic, exceptions are caught but recovery guidance is minimal. The server uses tool annotations appropriately (readOnlyHint), which is good. Input validation exists for SQL and pagination parameters but lacks detailed constraint documentation in descriptions. Search tools construct SQL internally (search_logs builds match_all query), which is pragmatic but the SQL construction lacks explanation in the parameter docs. Overall definition quality is solid but inconsistent, good naming and basic structure, but shallow descriptions and missing output documentation hold it back from B grade.

• Add output schema documentation to each tool. E.g. for search_sql: 'Returns an object with {success: bool, rows: [{field1, field2, ...}], total: int, truncated: bool}'. This must match the _apply_max_chars and result structure.
• Expand parameter descriptions with constraints. E.g. 'hours' → 'Lookback window in hours (must be positive, e.g. 1-744 for up to 31 days)', 'size' → 'Max results (1-1000, default 100)', 'interval' → 'Histogram interval (valid: 1min, 5min, 1hour, 1day, default 1 hour)'.
• Add discovery hints to search_logs, get_log_volume, and get_stream_schema. E.g. 'stream parameter: Name of the stream to query. Call list_streams first to see available streams.'
• Expand tool descriptions to 50-100 chars, including WHEN to use. E.g. search_sql → 'Execute SQL queries against OpenObserve logs. Use this for complex filters, aggregations, and time-range queries. Supports SQL with match_all() for full-text search. Results are capped at 1000 rows or 50KB.'
• Document what stream defaults to 'default' mean and whether users should override it. Either enumerate valid stream names or add a note that the default is used if stream parameter is not provided.
• Enhance error messages to include recovery guidance. E.g. catch stream-not-found errors and return 'Stream "xyz" not found. Available streams: [list]. Call list_streams() to discover.'
• Document truncation behavior in search_sql and search_logs descriptions. Mention max_rows and max_chars limits so LLMs understand why results might be capped.