Reviewing a pull request with Claude in Hyperpowers costs whatever your own Claude Code uses. On a small sample pull request, explaining it took about a minute and cost $0.21, a Claude review of all four areas cost $0.37 and took 43 seconds, and one area cost $0.11 and took 26 seconds. The extension itself is free.
Those are examples from one pull request, not a price list. This post shows what each step runs, what makes it cost more, and how to keep it down.
Who pays for the Claude usage?
You do, through your own Claude Code. Hyperpowers runs claude on your machine, the same as any other Claude Code session. On a Claude subscription, the usage counts against your plan. With an API key, it is billed per token at the model's price.
The extension needs no account, and there is no Hyperpowers server in between. It checks hyperpowers.dev for updates.
What does a review with Claude cost?
These are measured on a small sample pull request, with Claude Code's default model at the time:
| What ran | Time | Cost |
|---|---|---|
| Review with Claude, all four areas | 43 seconds | $0.37 |
| Review with Claude, the frontend area alone | 26 seconds | $0.11 |
Each area is one call to Claude Code with no tools and no MCP servers. It reads the diffs and the notes, up to 60,000 characters of diff per call, and up to three areas run at once. A bigger pull request, or a different model, gives different numbers.
What does explaining the pull request cost?
Create Walkthrough runs before any of that, to write the notes. On the same small sample pull request it took about a minute and cost $0.21. Here is what it runs:
- A small change is one call. By default it uses
claude-opus-5at medium effort. - A bigger change is split into parts of up to eight files and about 20,000 characters of diff. The parts run six at a time on
claude-sonnet-5at low effort, and one final call onclaude-opus-5writes the summary from the parts' notes.
A small pull request takes about a minute and a big one a few minutes. The hyperpowers.walkthrough.explainModel setting picks one model for every call instead.
What makes it cost more?
- The size of the diff. More changed lines means more to read, and a bigger change means more parts.
- How many areas you review. Each area in Review with Claude is its own call. A full-stack change has more areas than a frontend-only one.
- The model. A larger model costs more per token.
hyperpowers.walkthrough.reviewModelsets the model for Review with Claude. Empty means Claude Code's default. - Running it again. Review again runs every file reviewed so far fresh.
How do you keep the cost down?
- Review one area, or one file. Review with Claude offers This file, This area and Files I haven't read yet as well as All. Nothing runs until you pick one.
- Generated files are left out. Lockfiles, minified or bundled output, snapshots, and files under
dist,build,.next,outandgeneratedstay in the walkthrough, but Claude does not review them. - Results are kept per commit. What a review found, and what you decided, is kept for that commit. Reopening the pull request shows it again without running Claude.
- New commits only explain what changed. Update explains only the files the new commits changed and keeps the notes of every other file. After an update, Review with Claude offers The changed files.
- Finished parts are kept. When explaining a big change, finished parts are saved on disk, so running it again asks only for what is missing.
- Stop keeps what is done. Stop a running review and the areas that finished keep their findings.
Is it worth it?
That depends on the size of your pull requests and what your review time is worth. The model writes notes and drafts comments. You still read the code and decide, so the time it saves is the time you would spend working out what the change does.
Try it
Try Hyperpowers in your browser on an example pull request, or install the extension and follow Your first review. How to review large GitHub pull requests in VS Code covers the whole review.
Common questions
Is Hyperpowers free?
Yes. The extension is free. The Claude usage runs on your own Claude Code plan or API key.
Can I read a walkthrough without paying for Claude?
Yes. Reading a walkthrough someone exported needs only the extension. Creating one, and Review with Claude, need Claude Code.
Does Ask Claude about this change cost anything?
It runs on your Claude Code like the rest, and it reads only that one change, so it is a small call.
Are these numbers a guarantee?
No. They are examples from one small sample pull request with the default model at the time. Your diff and your model decide the cost.