Open-sourced MCP memory server for coding agents uses 3-pass RAG retrieval
Longjumping-Koala396 · reddit · 2026-07-26
Open-sourced a persistent memory MCP server for agentic coding tools, built around a 3-pass retrieval pipeline: vector search, query expansion, and LLM reranking.
Key details:
- Pass 1: cosine similarity with a 0.50 threshold
- Pass 2: LLM rewrites the query and retries at a lower global threshold when recall is thin
- Pass 3: an LLM grades and reorders all candidates before they are returned to the agent
- Exposed as MCP tools: remember, recall, forget, listmemories
- Self-hosted on Supabase/pgvector with per-project tags and row-level security
The author says it was designed for AI coding agents, but the memory layer is generic enough for other agent use cases too.
More from coding & agent
- It takes 10 messages to validate, 100 to ship an MVP, and 1,000 for production — tristanbob · 2026-07-26
- LiveKit adds huddles, letting humans and agents join the same live conversation — jasonkneen · 2026-07-26
- A coding agent deletes 50 lines, adds 1,292, then admits the real fix was tiny — xeophon · 2026-07-26
- Agent deck generation needs strict schemas and a deterministic fallback, not a magic API call — Live-Purpose-641 · 2026-07-26
- AI coding benchmarks should report performance, cost, and time distributions — zainhas · 2026-07-26
- Official Obsidian CLI guide shows how AI agents can drive a vault from the terminal — dSebastien · 2026-07-26