I built an augmented version of ChatGPT/Claude-style LLM interfaces, designed for power users like me. For the past week, I used it exclusively and have been quite pleased. It’s called interface0.

Today, interface0 is missing some of the polish of those canonical interfaces, and also some of the functionality that I personally find less important — but it makes up for it with several key new features. Most importantly:

  • all models in one place: you can use models from ChatGPT, Anthropic, Grok, Gemini, or anything else — even mixed in the same chat.
  • cross-provider memory: memories persist across those providers. I was starting to feel stuck with ChatGPT because it had built up memories, and found it hard to switch to blank-slate Claude 4 when it came out. But now, I’ve dumped all those memories into interface0 and am free to use any provider without lock-in.
  • swappable system prompts: you can create a range of system prompts/custom instructions and swap between them on a per-message basis. Sometimes I want the agent to be casual and concise, sometimes I want it to act as a lawyer, sometimes I want it to follow a certain template response. Rather than having default system prompts and pasting these further (and sometimes-conflicting) instructions into the user prompt, I now just swap the system prompt.
  • plus: inbox vs. all chats (separating unread/priority/pinned chats from others), outbound phone calls, and outbound + inbound emails (letting the agent interact with the world), forking chats, chained prompts (where a second message auto-triggers after the first one), and more…

Much more to come as well, both on polish and new features. I’ll continue to post about interface0 as it evolves, including demos.

interface0 is forked off of Julien Thibeaut’s excellent Zola project, which gave me the UI framework and scaffolding and some really slick touches, and then I’ve built these differentiated features on top of it.

You can sign up with your Google account to get on the waitlist, or find an invite code somewhere…

It’s very much in early beta. But once you try it out, I’d love your feedback. And if you have any ideas for new features or improvements, I’m all ears.

Looking for more to read?

Want to hear about new essays? Subscribe to my roughly-monthly newsletter recapping my recent writing and things I'm enjoying:

And I'd love to hear from you directly: andy@andybromberg.com