This one pulls your downloaded X/Twitter archive into a local SQLite database and exposes search, profiling, and draft generation through MCP tools. Instead of uploading a ZIP to ChatGPT as a one-shot context dump, you get repeatable hybrid retrieval, keyword and semantic search, and citation-grounded outputs that preserve tweet IDs like T1. Claude can query specific slices of your history without receiving the entire archive every time. The author built it around a read-only MCP interface, so your AI client retrieves evidence but never posts or calls external models unless you explicitly configure OpenAI-compatible endpoints. Useful when you want your assistant to draft in your voice while showing which old tweets informed the response.