Long-Term Memory
for AI Agents.
EvolutionDB is an open-source SQL database engine written in C, purpose-built to be the long-term memory for AI agents — relational storage, agent memory primitives, and a native vector index in a single binary.
MEMORY STORE ON
Starting EvolutionDB...
PG :5433 · EVO :9967
Why EvolutionDB?
Memory In-Engine
First-class MEMORY STORE and ENTITY STORE SQL primitives. Agent long-term memory is a storage-engine feature, not a bolt-on.
Native Vectors
A VECTOR(N) type and an HNSW index live in the core, enabling semantic plus lexical hybrid search without extra services.
Time-Travel
An MVCC engine with AS OF historical queries. Read your data as it was at any point in time — never lose the past.
Built for Agents, Grounded in Real Storage
One server speaks the PostgreSQL wire protocol and a native EVO text protocol, so it works with psql and DBeaver — and it ships an MCP server and a standalone agent for AI-native workflows.
Dual Protocol
PostgreSQL wire on :5433 and a native EVO protocol on :9967 from one binary. Zero-Docker embedded mode included.
Agent-Native
An MCP server and a standalone agent. Actions stay human-approved — drafts queue and a human approves before anything is sent.
Unified Storage
4KB slotted pages, B+ tree indexes, a Clock-Sweep buffer pool, and a write-ahead log with crash recovery.
Secure & Multi-Tenant
Per-tenant DATABASE, USER and GRANT isolation, with TDE page encryption (AES-256) and PBKDF2-SHA256 auth.
Give your agents a memory that lasts.
Open source under the MIT license. Clone it, run a single binary, and connect with the tools you already use.