Agentmemory is an MCP server that gives Claude Code, Codex CLI, Cursor, Gemini CLI, and Hermes persistent memory across sessions. It runs silently in the background, captures what your agent does, compresses it with AI, and injects the relevant context when the next session starts. No manual save, no fiddling with CLAUDE.md.
## The numbers
5,800+ GitHub stars within days. Hit #2 on daily GitHub trending on May 13. Apache 2.0, version 0.9.9 shipped May 11, install with one npx command. All data stays local on your machine.
## What’s inside
51 MCP tools, 6 lifecycle hooks (SessionStart, UserPromptSubmit, PreToolUse, PostToolUse, PreCompact, Stop), and 4 slash commands (/recall, /remember, /session-history, /forget). At 240 observations the project benchmarks at 1,900 tokens to maintain context versus 22,000+ for the built-in CLAUDE.md approach — a 92% reduction.
## Why it matters
The real bottleneck for long-running coding agents is context, not raw intelligence. Most fixes try to stuff more tokens; agentmemory compresses smarter. That’s why “persistent memory for coding agents” went from niche concept to top-5 trending in under a week. If you’re chaining Claude Code sessions across large projects, this is the thing to test.
Top AI Product 
Leave a comment