nanoclaw-repl

Run a session-aware Gemini REPL with persistent context by using a lightweight markdown-backed session store. This allows you to continue iterative assistant workflows exactly where you left off without losing previous command history or model settings.

Yes, you can switch models and load new skills dynamically during a REPL session. Use the /model command to change your current model and the /load command to add or update capabilities on the fly as you iterate on complex prompts and tooling.

The best way to manage risk when branching REPL workflows is to use the /branch command before making risky changes. This creates a safe separation in your session, allowing you to explore alternative paths without corrupting your original persistent context.

No, you do not need external dependencies to run a deterministic local command REPL. The system requires zero external runtime dependencies, ensuring reliable local command handling and markdown-as-database compatibility for session persistence entirely on your machine.

To search across past sessions and export a clean final run, use the /search command to find information across prior explorations and /compact to maintain clarity. Finally, use /export to share your results while tracking outcomes with /metrics.