macrocosm-os/macrocosmos-mcp

⏺ Summary

Adds Gravity API tools for large-scale data collection to the Macrocosmos MCP server. This enables users to collect unlimited social media data over 7 days, compared to the existing on-demand queries which are limited to 1000 results.

Changes

New Tools Added

• create_gravity_task - Start large-scale data collection from X, Reddit, or YouTube (7-day collection period)
• get_gravity_task_status - Monitor task progress and see records/bytes collected
• build_dataset - Build dataset early and get Parquet download links
• get_dataset_status - Check build progress and retrieve download URLs
• cancel_gravity_task - Stop a running collection task
• cancel_dataset - Cancel build or purge completed dataset

Improvements to Existing Tools

• query_on_demand_data: Clarified that @ symbol is optional for X usernames

Documentation Updates

• Updated README with complete documentation for all 7 tools
• Added parameter tables, example prompts, and workflow guides
• Documented that X topics MUST start with # or $ for Gravity tasks
• Added warning that build_dataset stops the crawler

Test Plan

• Created a Gravity task for #Bittensor data collection
• Monitored task status and verified records_collected updates
• Built dataset successfully (10,542 rows)
• Downloaded Parquet file and verified data integrity
• Verified all tools compile without errors