Four tools with complete schemas and descriptions. Naming follows verb_noun pattern (list_runs, load_run, query_logs_tool, get_node_summary_tool). Descriptions are detailed (100-300 chars) and explain WHEN to use each tool. Parameters have types and descriptions. However, query_logs_tool and get_node_summary_tool have awkward suffixes ('_tool'), and output schemas are not formally documented, only inferred from code. Error handling is minimal (no recovery guidance). No tool annotations (readOnlyHint, etc.). Pagination support is present but not consistently applied.

• Rename query_logs_tool → query_logs and get_node_summary_tool → get_node_summary to follow verb_noun convention and remove redundant '_tool' suffix.
• Add formal output schema documentation to each tool docstring. Example: 'Returns: {"run_id": str, "nodes": list[str], "severity_counts": dict, "time_range": [float, float], "num_entries": int}'.
• Add readOnlyHint=true to all four tools in their schema definitions to signal to LLMs that these are safe, non-destructive operations.
• Enhance error messages with recovery guidance. E.g., 'Node "sensor_driver" not found. Available nodes: ["motion_planner", "controller"]. Try get_node_summary_tool with one of these names.'
• Add numeric bounds to 'context' parameter: 'context (int, 0 - 100): Include N messages before/after each match. Defaults to 0 (no context). Max 100 to prevent excessive output.'