oadank/agent-wiki-mcp

Agent Wiki MCP is a local knowledge base that combines PostgreSQL + pgvector with BGE embeddings to provide a multi-agent shared wiki accessible via MCP (Claude Code, Codex, Hermes, Cursor, VS Code Copilot, OpenClaw). It replaced an embedded LanceDB design to gain stronger concurrency for multi-AI use. Who uses it: AI agents and developer teams that need shared memory, project handoff and programmatic access to documentation and project state. Key features: hybrid search (grep exact + pgvector semantic), multi-format ingest (PDF/DOCX/XLSX/Markdown), incremental vector indexing, shared memory for user prefs and project decisions, multi-AI concurrency, automated project-tracking (change detection, periodic sync, wake reminders) implemented without LLMs, MCP command set and scripts for querying/ingesting/index maintenance, and a Postgres schema (wiki_vectors with embedding vector(512)). Configuration uses a BGE embedding endpoint and standard PG env vars.

AI-related keywords found in this repo's description, topics, or README summary — grouped by category. Each badge links to the corresponding ranking detail page.