Vibe Coding DiscoverOri-Mnemos
View on GitHubLocal-first persistent agentic memory powered by Recursive Memory Harness (RMH). Open source must win.
Local-first persistent memory infrastructure for AI agents: an MCP server + CLI storing memory as markdown plus a SQLite index. Features a wiki-link knowledge graph, ACT-R decay, Hebbian co-occurrence edges and four-signal retrieval fusion, with lifecycle bridges for Claude Code, Cursor and other clients.
Use Cases
Give AI agents persistent long-term memory across sessionsShare one memory vault across Claude Code, Cursor, Codex and other MCP clientsMulti-hop retrieval and recursive query decomposition over markdown notesSemantic and BM25 search over a personal knowledge vaultTrack cognitive decay and prune stale notes automaticallyAnalyze knowledge graph structure (PageRank, communities, orphans)Capture and promote session insights into a versioned note vaultRun zero-cloud, no-API-key memory retrieval locally via MCP tools
Built With
- Language
- TypeScript
- Frameworks
- Model Context Protocol SDK · Hugging Face Transformers · better-sqlite3 · graphology · graphology-communities-louvain · Commander · Vitest · TypeScript · Node.js
Tags
agent-memory · persistent-memory · knowledge-graph · mcp-server · local-first · markdown · sqlite · vector-search · semantic-search · embeddings · memory-decay · spreading-activation · pagerank · cli · rag · recursive-retrieval