MemPalace is a free, open-source memory system that gives AI assistants and agents a persistent, searchable long-term memory. The project describes itself as the best-benchmarked open-source AI memory system, and it has attracted more than 59,000 stars on GitHub, with its latest updates in 2026. Instead of every conversation starting from zero, MemPalace lets a model remember what matters across sessions.
Built in Python on top of ChromaDB, MemPalace stores conversations, facts and documents as embeddings in a local vector store, then retrieves the most relevant memories when a new conversation needs them. It plugs into agent workflows through an MCP server, so tools like Claude Code and other MCP clients can read and write memory through standard tool calls. Because the store lives on your own machine, memory stays private and portable โ no hosted memory service sitting between you and your data. Integration is deliberately small: point the MCP server at a memory store and the host application gains recall across restarts, with no changes to the model itself. Because memories are stored as ordinary local data, they can be inspected, backed up and moved between machines like any other project files.
MemPalace is completely free and open source under the MIT license. There is no paid tier and no hosted service to subscribe to โ you run it yourself, on your own infrastructure, with your own models for embeddings.
MemPalace is for agent builders and heavy AI users who are tired of re-explaining context: developers adding memory to copilots, researchers benchmarking memory architectures, and anyone building assistants that should genuinely remember them over time.