Mem extends coding agents with beyond-Git memory — auto-capturing prompts, agent plans, and code changes as bound context. As your coding partner, it accelerates debugging, shares context in real time, reuses edits, prevents agentic infinite loops, and turns history into learning.
- 💬 Join our Discord and dive into smarter context engineering
- 🌐 Visit memov.im to visualize your Mem history and supercharge existing GitHub repos
- 📒 Context-bound memory: Automatically track user prompts, agent plans, and code changes — independent of Git history
- 🐞 Context-aware debugging: Isolate faulty context and leverage it across LLMs for 5× faster fixing
- 🤝 Team context sharing: Real-time alignment with zero friction
- ♻️ Change reuse: Reapply past code edits by description to save tokens when iterating on a feature
- 🛑 Loop guard: Prevent runaway agent auto-generation by intervening and halting infinite loops
- 🔍 History-driven optimization: Use past records and failed generations as reference context to boost future outputs
Please see docs/installation.md for detailed installation instructions.
Please see docs/installation_for_dev.md for detailed installation instructions.
These are available to MCP clients through the server:
-
mem_snap(files_changed: str)- Create a mem snapshot tied to the previously set user prompt. Handles untracked vs modified files intelligently. Argument is a comma-separated list of relative paths.
-
GET /health- Returns "OK". Useful for IDE/agent readiness checks.
MIT License. See LICENSE.