MCP Tools Reference

The main interface through which AI agents query and update the local Spoke node.

• brain_context: Retrieves a dense 8000-character project summary to bootstrap new chat sessions.
• brain_query: FTS5 full-text keyword search.
• brain_search: Semantic vector RAG queries.
• brain_recall: Fetches a specific memory block using its 4-character hex hash (e.g. 7c6a).
• brain_sources: Lists source files (PDFs, text files, code) feeding the RAG engine.
• brain_clusters: Performs unsupervised clustering to group related memories.
• brain_status: Reports health metrics and Master Hub connectivity.

• brain_store: Stores a structured memory block.
• brain_log_session: Packages and saves chat history into a queryable session log.
• brain_ingest: Parses and embeds directories/files (txt, md, pdf) into the RAG index.
• brain_snapshot_file: Generates and caches code summaries to minimize prompt token overhead.
• brain_sync: Runs bidirectional sync between SQLite database and local Markdown vault files.