rohitg00 / agentmemory

2026-02-25 · Source: Github Trending: All languages · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Robotics & Autonomous Systems · Depth: Intermediate, extended

Summary

Agentmemory is a persistent memory solution for AI coding agents, built on the iii engine, designed to eliminate repetitive explanations across coding sessions. It supports various agents like Claude Code, Cursor, Gemini CLI, and Codex CLI, integrating via hooks, MCP, or REST API. The system automatically captures agent actions, compresses them into searchable memory, and injects relevant context into new sessions. Benchmarks show agentmemory achieves 95.2% retrieval R@5 accuracy on LongMemEval-S, uses 92% fewer tokens (approximately 170K tokens/$10 per year) compared to full context pasting, and offers a real-time viewer on port 3113. It features a 4-tier memory consolidation pipeline, hybrid search (BM25 + vector + knowledge graph), auto-forgetting, and privacy filters.

Key takeaway

For AI Architects designing or implementing AI coding agent workflows, agentmemory offers a critical solution to overcome the "forgetting" problem. By integrating this persistent memory engine, you can significantly reduce token costs and improve agent efficiency, allowing agents to retain context across sessions. Evaluate its hybrid search and multi-agent compatibility to enhance your development environment, ensuring your agents build knowledge over time without constant re-explanation.

Key insights

Agentmemory provides persistent, searchable memory for AI coding agents, enhancing efficiency and reducing redundant explanations.

Principles

• Automate memory capture via agent hooks.
• Employ hybrid search for robust retrieval.
• Consolidate memories across multiple tiers.

Method

Agentmemory captures tool use via hooks, deduplicates, privacy filters, compresses with LLMs, embeds, and indexes. Sessions are summarized, and knowledge graphs extracted, then used to inject relevant context into new sessions via hybrid search and token budgeting.

In practice

• Install agentmemory globally for persistent memory.
• Use `agentmemory connect [agent]` to wire your coding agent.
• Monitor memory building live via the web viewer on port 3113.

Topics

• AI Coding Agents
• Persistent Memory
• Hybrid Search
• Memory Consolidation
• iii Engine

Code references

• rohitg00/agentmemory
• iii-hq/iii
• openai/codex
• tinyhumansai/openhuman
• google-gemini/gemini-cli

Best for: AI Architect, AI Engineer, Machine Learning Engineer, Software Engineer

Related on AIssential

• Agent memory resolved? — A git-like memory framework significantly enhances coding agent performance and context persistence across sessions.
• How xMemory cuts token costs and context bloat in AI agents — xMemory uses a hierarchical memory structure and uncertainty-gated retrieval to optimize long-term AI agent coherence and reduce token costs.
• AI Agent Amnesia? Here’s the Open-Source Fix That Works. — Hermes Agent uses SQLite FTS5 and LLM summarization for persistent, cross-session AI memory.
• A semantic memory layer for local AI agents — no vector DB, one file, runs on CPU — A lightweight, local semantic memory solution for AI agents avoids complex vector databases for smaller datasets.
• New agentic memory framework uses 118K tokens per query. LangMem burns through 3.26M. — MRAgent dynamically reconstructs memory during reasoning, significantly reducing token use and improving long-horizon task performance.
• Engineering Memory for AI Agents: A Practical Beginner’s Guide — AI agent memory selectively retains past interaction details to enable contextual, multi-step task execution.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Github Trending: All languages.