Self-hosted hybrid agent memory on Qdrant combining vector, BM25, temporal and optional graph recall; via MCP, HTTP or Python.
Setup
This project documents no single launch command; follow its repository instructions.
Mnemostack
udjin-labs
Related
More- PPenfield MCPPenfieldOfficialPersistent memory for agents with hybrid BM25, vector and graph search, context checkpoints and artifact storage.Knowledge & MemoryRemote only6 Repository stars
- Mmem0 MCP Self-HostedelvismdevSelf-hosted mem0 memory for Claude Code using Qdrant vector search, Neo4j knowledge graph and Ollama embeddings for cross-session memory.Knowledge & Memory111 Repository stars
- LLociIvenKooLabShared local memory for AI coding tools with hybrid search over notes, docs and PDFs and cited answers, via MCP and HTTP.Knowledge & Memory94 Repository stars
- FFEGISp-funkDefines agent tools in YAML schemas and stores every tool call in Qdrant for semantic search and memory across sessions.Knowledge & Memory24 Repository stars
- CCeliums MemoryCeliumsPersistent memory engine for agents with a journal, per-user isolation and hybrid PostgreSQL/Qdrant search.Knowledge & Memory23 Repository stars
- RRecallNestAliceLJYLocal-first memory for coding agents: hybrid vector/BM25 retrieval with reranking, knowledge graph, session checkpoint/resume and scope isolation.Knowledge & Memory15 Repository stars