Mcp External Memory

Gives Claude persistent memory through semantic search over stored text snippets. Exposes seven tools: memory_store for saving content with optional namespaces and tags, memory_search for cosine similarity lookups, plus get, delete, list, update, and stats operations. Supports three embedding backends out of the box: TF-IDF with no dependencies, OpenAI's text-embedding-3-small, or local Ollama models. Reach for this when you need Claude to remember user preferences, project context, or domain knowledge across conversations. Stores everything in a local database that persists between sessions. Install via pip, configure the embedding backend through environment variables, and start storing queryable memories immediately.