Mem-Dog

Mem-Dog is an open-source AI memory platform that ingests data from 300+ apps and turns it into searchable, private knowledge on your own infrastructure. Built for developers and teams building AI applications, it pulls from sources like Slack, Gmail, GitHub, WhatsApp, HubSpot, and Salesforce through Nango OAuth, then runs the data through 42 AI agents that classify it, extract entities, generate embeddings, and build knowledge graphs. Retrieval spans five search modes: vector (pgvector), full-text (BM25), hybrid, graph-based (Neo4j), and merged signals, with RAG chat that returns answers backed by numbered inline citations. Mem-Dog supports 10 memory types (timeline, session, episodic, semantic, factual, organizational, and more) and a temporal knowledge graph that tracks facts with valid_at/invalid_at timestamps for point-in-time queries. It ships SDKs for Python, TypeScript, Go, Rust, and Ruby (around 120 methods each), an MCP server for Claude Desktop and Cursor, and adapters for LangChain, CrewAI, and OpenAI. Because it is self-hosted and open source under Apache 2.0, no data leaves your environment; it runs on any cloud (GKE) or even a Mac mini. Pricing is free with no per-seat or usage fees.

Developers and teams building AI applications use Mem-Dog to ingest data from 300+ apps into a private, self-hosted memory layer with semantic search and cited RAG chat.