Agent Memory
Self-hosted memory for AI agents over MCP: knowledge graph, vectors, graph RAG, cited answers.
Semurg is a local persistent memory for AI agents over the Model Context Protocol. Semurg provides seven memory tools over MCP and the same seven calls over plain HTTP. Semurg uses a knowledge graph with no schema to write.
Each answer carries a citation with the document, timestamp, and quoted sentence. Semurg can search by meaning locally with its embedding model installed. Semurg provides separate memories through named scopes.
Semurg runs on x86_64 machines: Ubuntu 24.04 directly, Windows through WSL2, or Docker. One node is free with no licence key.
Summary drafted from the project's own website. Every sentence is backed by text on that page and was reviewed before publishing.