I may be missing something but in my attempt to use tidewave for helping with my Phoenix LiveView UI (I have used the MCP successfully with Claude Code and Zed for backend work).

Is there a way to start new chat(s) that can continue from previous work. In the command line claude I choose the pass the -c flag if I break off, review, do my own work etc. IF I want claude to jump back in. This approach together with having Claude add any findings to the CLAUDE.md file if they are useful, saves a lot of time, tokens and money jumping back in to some tasks.

I ask this question because in my tidewave experiment on the web UI, often when Claude needs to test something out, it goes around the same loops several times even after I have told it what to do in prior chats. This results in a lot of tokens being generated just getting it back up to speed on the application. I have asked Claude to write findings to AGENTS.md but that made no difference.

Thanks in advance for any tips, hints, help or simply pointing out my mistakes.

Willy