Nemp Memory

Nemp
  Agentic memory that evolves with your work
  Keep the decisions, preferences and context your next session needs.Capture what matters. Find it again. Refine it as your work changes.

  Get started · Commands · How it works · Roadmap

Your work moves forward. Your memory should too.

A decision changes. A workaround becomes a rule. Yesterday's assumption stops being true. The next agent session needs the context you still trust, not another copy of an old conversation.

Nemp gives that context a home in local, editable files. Save a decision, retrieve it in a later session, and update the same memory when the project changes. Keep project knowledge in .nemp/ and reusable preferences in ~/.nemp/.

Nemp Open v0.3.0 is the free, MIT-licensed foundation: agent-guided commands, keyword-based retrieval and CLAUDE.md sync through the Claude Code integration. The host agent follows Nemp's instructions; Nemp is not a separate background intelligence service.

Start with one useful memory

Run these inside Claude Code, from your project directory:

Restart Claude Code if the new commands aren't available, then:

When the decision changes, save the same key again:

That is the core loop: save useful context, use it, keep it current.

Want a head start? /nemp:init detects a JavaScript or TypeScript stack from package.json. /nemp:auto-sync on enables CLAUDE.md updates after supported memory changes. Full setup guide →

A memory you can work with

“Evolves” means memory can change alongside the work. Today, this loop is agent-guided and user-directed. Automatic learning, validated corrections and reversible archival are future work.

Built for the things worth carrying forward

- Decisions with a reason. Keep the constraint behind a choice, not just the chosen tool
- Project…