Skip to content

[GitHub Trending] rohitg00/agentmemory

9.2 relevance
Score Breakdown
technical depth
8
novelty
8
actionability
9
community
8
strategic
6
personal
10

Scored daily by a customisable AI persona to surface the most relevant engineering leadership news.

Persistent memory for AI coding agents, directly addresses agent orchestration needs.

2026-05-12 ai/ml GitHub Trending
#1 Persistent memory for AI coding agents based on real-world benchmarks - rohitg00/agentmemory
Summary

agentmemory is an open-source persistent memory engine for coding agents (Claude Code, Cursor, Gemini CLI, etc.) that eliminates re-explaining context across sessions via MCP, hooks, or REST API. It automatically captures agent actions, compresses them into searchable memory using BM25 + vector + graph search (RRF fusion), achieving 95.2% R@5 on LongMemEval-S—outperforming mem0 (68.5%) and Letta (83.2%)—while consuming only ~1,900 tokens/session ($10/yr with local all-MiniLM-L6-v2 embeddings). Built on iii engine with SQLite, no external dependencies, and v0.9.0 adds a filesystem connector and audit policy.

Key Takeaway

Integrate agentmemory as a shared memory server for all your MCP-compatible coding agents to eliminate session context loss and reduce token costs by 10x.

Why it matters

For a senior engineer building agent orchestration, this solves the critical problem of context loss across sessions, enabling agents to retain project-specific knowledge without manual re-teaching or expensive LLM summarization.