Talk to Claude Code. Force Yourself to Slow Down and Think

github.com

2 points·tmshapland··4 comments

Hi HN community,

Audiochatty gives you a way to talk to Claude Code. When Claude Code finishes a turn, it describes what it did to you. You can then tell it what to do next.

It works by wrapping Claude Code in a pseudo-terminal. audiochatty run opens a pseudo-terminal, starts claude inside it, and sits as a layer between you and the terminal running Claude Code.

I've linked to the plugin that shows more about how it works technically. If you want to try it out, there's a link in the readme to a hosted version of the frontend, backend, and voice agent that has lots of free minutes.

I've wanted to be able to talk with Claude Code for a while. (Anthropic will probably build it eventually, but I have grown impatient.) I skim Claude Code's jargon-rich output and don't take the time to think through what it did and what I want it to do next. By talking to Claude Code, it forces me to slow down, really understand what it did, ask follow up questions, and better define what I want it to do next. It's helped me write better code, better product requirement docs, better marketing plans, etc.

is this substantially different than /plan mode?

thank you for asking. we take the last message from a claude code turn, which is the summary of what it did, and hand it to a voice agent. you can then talk to the voice agent about what Claude Code did in its turn. It's helpful for me to have to stop and listen to a spoken description of what the agent did. It also gives me space for asking follow up questions.

Oh, cool. I missed the voice integration part. I think this is cool (and why I always work in /plan mode myself) gets me to really think and understand and challenge what is being proposed.

Reply on news.ycombinator.com