This is an early release preview. You may encounter bugs.
Human Review logo
Unclaimed

Tool interface

Human Review

A visual tool to edit HTML and Markdown files, leave comments like a Google Doc, and send feedback to your AI agent. Works inside your favor

B 69/100 GitHub score ? This grade is derived from GitHub signals, not user votes. Open for the full breakdown.
Discussion

01 / Summary

About Human Review.

Human Review opens an HTML or Markdown file, or a locally served page, in a browser, where you edit the text directly, leave comments anchored to an exact phrase or element, and send every change to a coding agent in one batch. It installs as a skill so an agent can open it on request, and the whole session runs on your own machine: no account, cloud service, database, or API key.

The alternative it replaces is describing edits in chat, then checking whether the agent understood each instruction. Here you make the change yourself where you can, comment where you cannot, and click Send once. The agent applies the batch to the source and refreshes the page for another pass.

How the agent waits depends on the harness. In Claude Code it waits in the background and picks the feedback up the moment you hit Send. In Codex and other agents it waits during its turn, and if that turn has already ended, a message prompts it to collect the feedback. Closing the tab or clicking End review releases the agent either way, and feedback you never sent is kept for the next time you open the page, where you can restore or discard it. Direct edits and resizes to an HTML file save automatically and are undone with Discard or Revert all. Markdown files and local pages need a Send so the agent can apply them to the source.

  • Direct text editing: change copy in place, with bold and italic formatting
  • Lists and links: type - or 1. to start a list, indent with Tab, and add or edit a link with ⌘K
  • Image handling: resize by dragging a corner, move by dragging, and paste images from the clipboard
  • Block rearrangement: hover a block and drag the handle on its left edge to move it elsewhere on the page
  • Anchored comments: select a phrase, or click an image, chart, or section, and attach a comment to it
  • Editable comments: reword a comment after writing it, including one already sent, and the new wording reaches the agent on the next Send
  • Batched feedback: one Send delivers every edit and comment at once instead of a long chat message
  • Local page review: point it at a running local URL to review a live page rather than a file
  • Multi-page review: Command-click links to move between pages without losing pending feedback
  • Draft retention: unsent feedback is preserved and can be restored or discarded the next time the page is opened
  • Local only: the CLI, review server, and browser interface all run on your computer
  • Agent instructions: a bundled skill file teaches Claude Code, Codex, and other agents how to drive the review session

02 / Discussion CREDIBILITY-GATED

Discussion

Reading is open to everyone. Posting and voting need a verified identity or a GitHub grade of B or higher.

  • No discussions yet.

04 / Build

Build with Human Review.

Browse the catalog for frameworks, tools, and harnesses, each scored on real GitHub credibility.

Get Human Review →

Browse the catalog