Introduction

mnestic is a general-purpose, transactional database that uses Datalog for queries, is embeddable but can also handle large amounts of data and concurrency, and treats graph and vector data as first-class. It is a maintained fork of CozoDB, continued under MPL-2.0 and tuned as a substrate for agentic memory.

What "embeddable" means

A database is embedded if it runs in the same process as your program, rather than as a separate server you connect to over a socket. SQLite is embedded; Postgres and MySQL are client-server. Embedded databases need no setup and run almost anywhere.

mnestic is embeddable rather than strictly embedded: you can run it in-process for zero-setup local use, or stand it up as an HTTP server when you want to share resources and handle more concurrency.

Why graphs

Data is inherently interconnected, and the most useful insights often live in graph structures several levels deep. Rather than forcing your data into a labelled-property graph model, mnestic keeps the relational model — an algebra that handles graph data cleanly and composes far better.

Why Datalog

Datalog expresses every relational query, and makes recursion easy to write, more powerful, and usually faster than the SQL equivalent. Queries compose piece by piece: you build a result from small named rules.

# a recursive rule: everyone reachable by following "follows" edges
reachable[to] := *follows{ from: $start, to }
reachable[to] := reachable[via], *follows{ from: via, to }
 
?[person] := reachable[person]

Relational, graph, and vector in one engine

The same query can join stored relations, recurse over graph structure, and search a vector index by meaning — no separate systems to keep in sync.

# 2-hop neighbours of a seed, re-ranked by embedding distance
recall[to] := *recalls{ from: $seed, to }
recall[to] := recall[via], *recalls{ from: via, to }
 
?[memory, dist] :=
    recall[memory],
    ~memory:embedding{ memory | query: $cue, k: 12, ef: 80, bind_distance: dist }
:order dist
:limit 12

Where to go next

• New to the engine? Start with the Tutorial.
• Installing it in a Rust project? See Installation.
• Want the headline fork features? Jump to What mnestic adds.
• Looking for a specific function or operator? See the Functions reference.