**Problem** Current large query results are sent entirely to LLMs, causing: - Token waste - Performance issues - No reusable outputs for downstream workflows **Solution** Add `export_to_json` tool that: 1. Streams results directly to filesystem 2. Returns only minimal metadata to LLM: ```json { "success": true, "file_path": "/exports/query_12345.json", "count": 15000, "sample": [first_3_docs] } ``` 3. Preserves data for: - Visualization tools - Subsequent MCP processing - Human verification