@Jmoon_174 - Headroom Shows Your Claude

Open original

@Jmoon_174 on X The author says CLAUDE.md solves the instruction side of working with Claude, but the other half is tracking how much context you have left before you hit the limit. They built Headroom, a one-line install that puts a percentage readout in your statusline.

  • CLAUDE.md handles persistent instructions for Claude, which is only half the problem.
  • The other half is knowing your remaining context budget before you run out.
  • Headroom installs in one line and displays context usage as a percentage in your statusline.

The whole post is a reply tweet, and it's short. Someone named jmoon_174 is responding to @kirillk_web3 about working with Claude, the AI model. Their point is that a CLAUDE.md file takes care of one problem, which is giving the model standing instructions so you don't repeat yourself every session. The second problem, they say, is tracking how much of the model's context window you've used before you run into the limit and lose work or get cut off. To address that second problem, they built a tool called headroom and posted a link to its GitHub repo under the henchmarketing-rgb account. The pitch is a one-line install that puts a percentage readout in your statusline, so you can see how full your context is while you work. The post doesn't give benchmarks, screenshots, or details on how the percentage is calculated, so you'd need to check the repo to see whether it reads actual token counts or estimates them. Why it matters: if you use Claude Code or similar tools for long sessions, running out of context mid-task is a real annoyance, and a persistent meter is a cheap fix. Whether headroom is well built is a question the tweet can't answer. It's worth a look at the repo's README and issues before installing anything, since it's a third-party tool that will sit in your shell's statusline.

Ask about this article