Stop Pasting Feedback Into Your Terminal: Plannotator Puts Agent Reviews in Your Browser
You've been there. Your AI coding agent spits out a plan or finishes a chunk of code, and now you're squinting at a wall of terminal text trying to figure out what to say about line 47. You type out a paragraph of feedback, hit enter, and hope the agent understood what you meant. There's a better way to do this, and it's called Plannotator.
What It Does
Plannotator is a local, browser-based review surface for AI coding agents. It's designed to plug directly into agents like Claude Code, Codex, Copilot CLI, Gemini CLI, OpenCode, Kiro, Droid, Amp, and Pi through their hooks and commands. When your agent proposes a plan, generates HTML, or finishes writing code, that work opens in your browser instead of staying buried in your terminal.
From there, you can annotate plans, specs, messages, and HTML artifacts directly. You mark things up, leave comments, and send feedback straight back to the agent to act on. There's also a full code review mode where you can review local changes or remote PRs, comment on diffs, and suggest code. The review side works with Git, GitButler, Jujutsu (jj), Perforce (p4), GitHub, and GitLab, so it's not picky about your version control setup.
The whole thing runs locally in your browser, which means you're not shipping your code or plans off to some third-party service just to leave a comment on a diff.
Why It's Cool
-
It meets your agent where it already works. Rather than asking you to adopt a whole new workflow, Plannotator hooks into the agents you're already using. The list is long—Claude Code, Codex, Copilot CLI, Gemini CLI, OpenCode, Kiro, Droid, Amp, and Pi—so there's a decent chance your tool of choice is covered.
-
Browser-based annotation is just better for review. Highlighting a specific paragraph in a plan and attaching a comment is far more precise than trying to describe what you mean in a chat message. The same goes for diffs—side-by-side views with a file tree beat scrolling through terminal output every time.
-
AI is built into the review process itself. You can ask AI about anything you're reviewing, or launch AI reviews that post comments directly to the diff. That's a nice touch—it means the tool isn't just a passive surface, it's actively helping you figure out what to say.
-
Broad version control support. Git is the obvious one, but seeing GitButler, Jujutsu, Perforce, GitHub, and GitLab all listed means this isn't a tool that only works for people on the most mainstream stack.
-
There's a terminal option too. If you'd rather stay in the terminal, Herdr Annotate brings Plannotator's functionality there. It's a separate project, but it's linked right from the README, so you've got options depending on how you like to work.
How to Try It
The README points to a full installation guide rather than dumping commands inline, so the best path is to head straight to the docs. Here's the short version:
- Check the installation guide for setup instructions specific to your agent.
- If you want to see it in action first, there's an original demo video and a plan review demo with Pi.
- Once installed, Plannotator hooks into your agent. When the agent proposes a plan or finishes code, it'll open in your browser for review.
- Annotate, comment, and send feedback back to the agent.
The repository is at github.com/backnotprop/plannotator. There's also an official site and a related project called Visual HTML Skills if you're working with HTML artifacts.
Final Thoughts
Plannotator solves a real friction point: reviewing AI-generated work in a terminal is clumsy, and the feedback loop between you and your agent deserves better tooling. The browser-based approach makes annotation and diff review feel natural, and the breadth of agent and version control support means it's not a niche solution. If you're regularly working with coding agents and find yourself wishing you could just point at a thing and say "fix this," Plannotator is worth a look. It's a practical tool for a workflow that's only going to get more common.
Follow @githubprojects for more developer tools and open source projects.