Persistent memory engine · v2.7.0

Your agent remembers, across every session.

Mneme is a local, encrypted memory server for AI agents. Facts, decisions, and context persist in SQLite on your machine: searchable by keyword and meaning, decaying by confidence, never leaving your control.

$ docker pull ghcr.io/perseus-computing-llc/mneme:2.7.0
View on GitHub ↗
43
MCP tools
AES-256
GCM encryption at rest
FTS5 + vec
Hybrid search
0
Cloud dependencies
What it does

Memory that's layered, searchable, and yours.

Hybrid recall

FTS5 keyword search and vector similarity in one query: exact matches and semantic neighbors, ranked together.

Encrypted at rest

AES-256-GCM on a local SQLite store. Your memory never leaves the machine: no cloud, no account, no telemetry.

Confidence decay

Memories age. Stale facts lose confidence over time and fresh ones surface first. Recall stays relevant, not cluttered.

43 MCP tools

A full Model Context Protocol surface (store, recall, link, forget, and reflect), usable by any MCP-aware assistant.

How memory works

One local store. Layered by recency and confidence.

Every memory is written to an encrypted SQLite file with an embedding and a confidence score. On recall, Mneme blends keyword and vector hits, weights them by freshness, and returns a compact, ranked set: the same shape your assistant already reads.

mneme recall "deploy"
 staging = eu-west-2/stg-a  conf 0.98
 deploy via ./scripts/deploy.sh  conf 0.95
 helm chart v0.4.1  conf 0.91
▸ old: helm v0.3  conf 0.22 (decayed)

 3 memories · 7ms · FTS5+vector
Compose, don't replace

Mneme remembers. Perseus compiles.

Point Perseus at your Mneme store and your assistant opens with both live environment state and durable memory, resolved into one byte-stable file. Two single-binary tools, one build-pair.

Explore Perseus ↗

Local-first memory in 60 seconds.

Install the server, point your assistant at it. Everything stays encrypted on your machine.

MIT licensed AES-256-GCM 0 cloud dependencies v2.7.0 NVIDIA Inception Member