Zero-dependency, SQLite-backed AI memory layer for agents, MCP-ready, local-first, and benchmarked against BEAM and LongMemEval.
Qué hace
Mnemosyne (AxDSan) is a zero-dependency, SQLite-backed AI memory layer for agent frameworks. Install with one pip command and get sub-millisecond recall, a built-in MCP server, and a 3-tier BEAM architecture, no cloud, no Docker, no external vector DB required.
Most agent memory solutions demand Postgres, Docker, or a SaaS subscription before a single fact gets stored. Mnemosyne collapses the whole stack into one SQLite file and one pip install, removing the infrastructure tax that makes persistent agent memory painful to ship.
Mnemosyne is a universal memory layer built for the Hermes agent framework but usable with any agent. Per the README, it supports 8+ platforms including Cursor, Claude Code, OpenAI Codex CLI, Windsurf, OpenWebUI, and OpenClaw, via MCP (stdio or SSE) or a direct Python SDK import. No external services required.
The core is BEAM (Bilevel Episodic-Associative Memory), a 3-tier system: working memory for hot context (TTL-based eviction), episodic memory for long-term storage (sqlite-vec + FTS5 hybrid search), and a TripleStore for temporal knowledge graphs with version chains. Hybrid scoring blends 50% vector similarity, 30% FTS5 rank, and 20% importance, all inside SQLite. Binary vector compression (MIB) shrinks 384-dim float32 embeddings to 48 bytes, a 32x reduction, using Hamming distance inside SQLite with no ANN index.
Per the README, Mnemosyne scores 98.9% Recall@All@5 on LongMemEval (ICLR 2025) and 65.2% on the BEAM end-to-end QA benchmark (ICLR 2026). Recall holds flat across dataset scales from 100K to 10M tokens. Episodic compression delivers 9.4x storage savings. The system reports 100% abstention accuracy: it never hallucinates on unknowns.
The Python API is minimal by design: remember() and recall() cover the main path. Advanced features include scoped global memories, expiry dates on facts, entity extraction, LLM-driven fact extraction, per-domain memory banks, and direct BEAM access for custom consolidation workflows.
The README benchmarks are self-reported, and recall@10 sits at 20% flat across scales in the BEAM retrieval table, a number worth understanding before assuming the system surfaces everything relevant in a given query. The [all] install flag adds vector search as an optional dependency, meaning basic usage trades retrieval quality for zero deps.
Mnemosyne is the sharpest option when you want persistent agent memory that deploys anywhere without infrastructure ceremony. MIT-licensed, local-first, no telemetry, and MCP-native out of the box, it fills the gap between raw vector databases and full managed memory services.
Características
Notas de campo
Revisado el Jun 26, 2026
Ideal para
Resultados para creadores
Ten cuidado
Probado con
Profundiza
HN thread explicitly discussing Mnemosyne as an Anki alternative, offering practitioner commentary on its spaced-repetition approach and community reception.
Mnemosyne is a zero-dependency AI memory layer that gives any agent persistent, searchable memory backed by a single SQLite file. It uses a 3-tier BEAM architecture (working memory, episodic memory, and a temporal TripleStore) to store and retrieve facts without requiring Postgres, Docker, or any external vector database. It ships with a built-in MCP server, a Python SDK, and integration guides for Cursor, Claude Code, Codex CLI, Windsurf, OpenWebUI, and OpenClaw.
Run `pip install mnemosyne-memory` for the base install, or `pip install "mnemosyne-memory[all]"` to include vector search and the full MCP server. For MCP-based tools like Cursor or Claude Code, add a one-line JSON block pointing to the `mnemosyne mcp` command in the platform's MCP config file. For Python agents, import `remember` and `recall` directly from the `mnemosyne` package and call them anywhere in your agent loop.
Yes, Mnemosyne is released under the MIT license with no paid tiers mentioned in the sources. It is fully self-hostable and requires no cloud account or subscription. The optional sync server can be run on your own infrastructure (Docker, bare metal, or Fly.io).
Mnemosyne is best for agent developers who need persistent cross-session memory without spinning up cloud infrastructure. It excels in local-first setups where all data must stay on-device, in MCP-native workflows (Cursor, Claude Code, Codex), and in scenarios where episodic compression and a temporal knowledge graph add real value. Per the README, it supports 8+ agent platforms out of the box.
Mnemosyne is the only option in the comparison table that is local-first, zero-dependency, has a built-in MCP server, and is MIT-licensed, all at once. mem0 requires Qdrant or Postgres and scored 49% on LongMemEval vs. Mnemosyne's 98.9%; Honcho requires Postgres plus three LLMs and is AGPL-licensed; ChromaDB is a vector database only and has no memory architecture. The main area where Mnemosyne trails is raw BEAM end-to-end QA, where Hindsight scores 73.4% vs. Mnemosyne's 65.2% at the 100K scale, per the README benchmark table.
The BEAM retrieval recall@10 is reported as 20% flat across all dataset scales in the README's retrieval table, which means the system may not surface every relevant memory in a single query pass. Benchmark figures are self-reported and should be treated as directional until independently replicated. Vector search is an optional extra (`[all]`), so the base install falls back to keyword-only retrieval. Finally, the README notes that users are solely responsible for the content they store, and sync encryption must be explicitly enabled.
Mnemosyne@mnemosyne_oss
“LongMemEval 98.9%. BEAM 65.2% SOTA. 17K+ PyPI downloads. 750+ GitHub stars. Zero-dependency, local-first. Self-hosted agent memory that actually works.”
MrRuSs3LL@mrru5s3ll
“Found a ripper memory layer for AI agents — Mnemosyne (AxDSan/mnemosyne, 959⭐). Zero deps, single SQLite file, sub-millisecond recall. Built for Hermes first but works everywhere: Cursor, Claude Code, Codex, OpenWebUI, OpenClaw, any MCP cli…”
Infraestructura de memoria para construir agentes de IA con estado, almacena mensajes y eventos, permite que Honcho razone en segundo plano y luego consulta representaciones de peers, contexto de sesión o insights en lenguaje natural desde cualquier modelo o framework.
Best for Dar a los agentes memoria a largo plazo por usuarioFree, open-source shadcn-style registry of production-ready AI agent recipes for developers who want full code ownership with zero config.
Best for Teams that want full code ownership over agent logic without being locked into an opaque runtime dependencyEl sistema de orquestación de código abierto de OpenAI para Codex, convierte el trabajo de un proyecto en ejecuciones de implementación aisladas y autónomas para que los equipos gestionen el trabajo en lugar de supervisar agentes de codificación.
Best for Gestionar el trabajo en lugar de supervisar agentes de codificaciónOpen-source OpenTelemetry observability platform with traces, logs, metrics, service maps, session replay, and an AI agent MCP server, for backend and platform teams.
Best for Teams wanting unified traces, logs, metrics, service maps, and Kubernetes monitoring in a single OTel-native toolAplicación de escritorio local-first y de código abierto para diseño agéntico; genera y edita interfaces web, móviles y prototipos de UI, presentaciones, imágenes y video, con enrutamiento bring-your-own-key entre los principales LLMs.
Best for Generar UI, presentaciones y contenido multimedia coherentes con la marca a partir de un DESIGN.mdUn escáner de seguridad de código abierto para skills de agentes de IA que detecta vulnerabilidades, patrones maliciosos y riesgos en la cadena de suministro a través de 64 patrones antes de instalar una skill, con análisis estático y evaluación semántica opcional mediante LLM.
Best for Verificar una skill de agente de IA antes de instalarla