Memory for AI agents
MenteDB gives your agent long-term memory through one call: process_turn. Each turn it stores what was said, retrieves what is relevant, and hands back the memories to put in your next prompt. That loop is the whole integration.
It is a purpose-built Rust storage engine, not a wrapper around a vector database. It combines HNSW vector search, a knowledge graph, bi-temporal validity, and transformer-aware context assembly so the memories your agent reads are clean, current, and ordered for attention.
Cloud
Managed. Get an mdb_ API key, call the REST endpoint, done. Fact extraction and the overnight maintenance run for you, no infrastructure and no LLM keys to wire up.
Self-host
Run the engine yourself with cargo install mentedb-server or Docker, or embed it in your process. You own the data, the LLM key, and the schedule. Nightly maintenance is built in.
Building your own agent? Jump to the for a runnable path, or for a full example in every combination of cloud or self-hosted and Python, Node, or Rust.
mente (Spanish): mind, intellect.