Waggle MCP

Waggle MCP is an open-source AI memory tool that gives large language models persistent, graph-structured memory through a Model Context Protocol (MCP) server. Instead of losing context between sessions, it stores decisions, facts, preferences, and their relationships as typed nodes and edges in a local knowledge graph, so AI agents can recall and build on prior work across projects. Waggle MCP uses hybrid retrieval that combines vector search, BM25 keyword matching, and graph traversal to surface the most relevant context, and it supports temporal queries for point-in-time lookups and conflict resolution across sessions. It is local-first by default, running on SQLite with no cloud calls or API keys required, and automatic ingestion hooks capture conversational turns to extract durable information. Installation is a one-liner via pipx, and a single setup command detects MCP clients and configures memory hooks; a doctor command verifies the install. Waggle MCP works across MCP-compatible clients including Claude, Cursor, Codex, and Gemini CLI, making it a portable persistent-memory layer for developers building long-running AI agents that need decision continuity.

Developers building long-running AI agents use Waggle MCP to give LLMs persistent, graph-structured memory that survives across sessions and MCP clients.