π§ Mnema β Long-term memory for AI via MCP Γ Vector DB. Pluggable backends, hybrid search, Auto Dream consolidation, CLI + REST API + SDK.
Give your AI agents persistent, searchable memory. Solve the context-window problem with MCP Γ Vector DB.
Mnema (ΞΌΞ½αΏΞΌΞ± β Greek for "memory") is an open-source Model Context Protocol server that gives language-model agents a long-term memory layer. Instead of stuffing every relevant fact into a single conversation (and paying for it in tokens, latency, and lost context), store durable facts once and recall them later β by meaning, not by keyword.
- π MCP-native β drop it into Claude Desktop, Claude Code, Cursor, Zed, Cline, Continue, Windsurf, ZCode, or any MCP-compatible client. - ποΈ Pluggable vector backends β ChromaDB (embedded, default), Qdrant (local or remote), or sqlite-vec (pure-SQLite, zero-dep). - π§ Pluggable embeddings β sentence-transformers (offline, default), OpenAI, or Ollama (local server). - π Hybrid search β combines semantic similarity + tag overlap + decay scoring into a single ranked score. - β³ Memory decay β a forgetting curve (recency Γ frequency Γ importance) so the store stays focused on what matters. - π Summarization β plans how to condense many memories into a few high-level ones; the calling AI executes the plan (Mnema never calls an LLM on its own). - π₯ Multi-user / multi-session β scope-based namespace isolation (user:alice, session:abc, agent:bot-1). - π§ Offline by default β local sentence-transformers embeddings; no API keys required to start. - π¦ Programmatic SDK β use Mnema from Python without standing up an MCP server. - π» CLI β mnema add, mnema recall, mnema statsβ¦ for terminal-first workflows. - π§ͺ Well-tested β 86 tests across pure-fun
From the project README.
Add the radar badge to your README β it shows your project was picked up by MCP Radar and links to this page:
[](https://mcp.liqiwa.com/s/mienetic--mnema.html)
Local-first, project-aware memory MCP server for Claude Code, Codex, and other AI agents, with Obsidian and an owner dashboard.
MaxFreedomPollard/engRAMAutomatic one-click install. Encrypted, fully offline vector memory for AI agents; takes over built-in agentic memory.
MuhittinYilmazer/akanaSelf-hosted, local-first AI assistant server: swappable LLM providers, review-gated memory, an encrypted vault, and wake-word voice. No accounts, no telemetry.
AdamRussak/self-docSelf-hosted documentation RAG pipeline that serves semantic search to LLM agents over the Model Context Protocol.
menot-you/n-memoryMemory your agent can trust β every answer carries its source; when it doesn't know, it says so. Hermetic, local, MCP.
axisrobo/mneme-openOpen-source client SDKs (Python, TypeScript, Go), CLI, API docs, and prebuilt binaries for the Mneme agent cognition runtime (memory, context, ontology, knowledge). Apache-2.0.
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.