mnemosyne-oss/mnemosyne

Health v2 · maintenance only

Flags: young

Adoption not part of the score

2798 stars · 230 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

Mnemosyne is a local-first, SQLite-backed memory layer for AI agents, exposing remember/recall APIs via a Python SDK and 30+ tools over MCP. It combines vector (sqlite-vec) and full-text (FTS5) hybrid retrieval with tiered memory architecture, requiring no cloud services or external dependencies.

Use cases

• give my ai agent persistent memory
• store and recall user preferences across chat sessions
• add long-term memory to claude code or cursor
• local agent memory without cloud services
• semantic search over agent conversation history
• mcp memory server for coding assistants
• remember facts across agent runs in python

When to choose

• you need persistent, private agent memory that works fully offline
• you want a single pip install with zero external services or Docker
• your agent speaks MCP (Claude Code, Cursor, Codex, Hermes, OpenWebUI)
• you need fast sub-100ms hybrid vector + keyword retrieval on a single SQLite file

When to avoid

• you need multi-user, horizontally scaled shared memory across servers
• you require a managed cloud memory service with team sync built in
• you need heavy-duty vector search over millions of high-dimensional embeddings
• your stack is not Python and does not support MCP

Facets

library · maturity active

vector-database database rag mcp search-engine sdk cli artificial-intelligence large-language-models developer-tools privacy self-hosted python cli cross-platform self-hosted agent-memory sqlite local-first long-term-memory mcp-server hermes-agent vector-search fts5 privacy-first zero-cloud ai-agents retrieval-augmented-generation

Member repositories

For agents

markdown · JSON · MCP: product_card(name="mnemosyne-oss/mnemosyne")