agentmemory is a simple-to-use memory library for AI agents, powered by ChromaDB and PostgreSQL. It provides clear functions — such as create_memory() and search_memory() — to store 'memories' as dictionaries, with optional metadata and embeddings, and then retrieve them via semantic search.

By default, agentmemory relies on a local ChromaDB instance. By setting the appropriate environment variable, you can switch to PostgreSQL by pointing to its connection string — ideal for moving from prototype to production. Recent versions add asynchronous management, clustering, and read/write optimizations, particularly on the PostgreSQL side.

Designed for developers building agents and AI pipelines, agentmemory serves as a persistent memory foundation. It is a technical building block: it integrates into your code rather than being used via a graphical interface.