Building an AI Rubric Grader as a Chrome Side Panel for Google Classroom

Grading written assignments is one of the most time-consuming parts of teaching. A class of 30 essays, each scored against a 4-criterion rubric, can easily eat an entire weekend. I wanted to see how much of that work a well-designed tool could take off a teacher's plate — without pulling them out of the platform they already use.

The result is ClassGrade Pilot, a Chrome extension that opens a side panel right inside Google Classroom and drafts rubric-based scores and feedback for each submission. Here's what I learned building it.

Why a side panel instead of a separate web app?

Teachers already live in Google Classroom. Every extra tab, export, or copy-paste step is friction that kills adoption. Chrome's sidePanel API (Manifest V3) turned out to be a great fit:

• The panel stays open next to the student's work while the teacher scrolls through it.
• It can be scoped to specific sites, so it only shows up where it's useful.
• It survives navigation between submissions, which makes batch grading feel natural.

A minimal setup looks like this:

{
  "manifest_version": 3,
  "name": "My Grading Assistant",
  "permissions": ["sidePanel", "storage"],
  "side_panel": {
    "default_path": "sidepanel.html"
  }
}

// background.js — open the panel when the toolbar icon is clicked
chrome.sidePanel
  .setPanelBehavior({ openPanelOnActionClick: true })
  .catch(console.error);

Rubrics are the real product

The AI part gets the attention, but the rubric is what makes the output trustworthy. A vague prompt like "grade this essay" produces vague, inconsistent results. A structured rubric — criteria, point levels, and a clear description of what each level looks like — gives the model something concrete to anchor to.

A few lessons:

1. Make criteria explicit. "Uses evidence to support claims (0–4 pts)" works far better than "Quality of argument."
2. Describe every level. The difference between a 2 and a 3 should be written down, not implied.
3. Ask for justification per criterion. Scores tied to quoted evidence from the student's text are much easier for a teacher to verify.
4. Let teachers save and reuse rubrics. Most teachers grade the same assignment types repeatedly, so saved rubrics save a lot of setup time.

Keep the teacher in the loop

The goal isn't to replace the teacher's judgment — it's to give them a solid first draft. Every score and comment the extension produces is editable before anything is returned to a student. In practice this means the teacher shifts from writing feedback to reviewing and adjusting it, which is dramatically faster.

Designing a fair free tier

Pricing an education tool is tricky: teachers often pay out of pocket, so they need to try it on real work before committing. I settled on a free plan with 30 student credits per month and 2 saved rubrics — enough to grade a full class set and see if it fits their workflow — and a Pro plan for teachers who grade at volume.

What I'd tell anyone building for teachers

• Integrate into existing tools rather than asking people to switch.
• Structure beats clever prompting — invest in the rubric model.
• Make every AI output transparent and editable.
• Respect student privacy and only process what's needed for the task.

If you're a teacher (or know one who's drowning in grading), you can try it at classgradepilot.com. And if you're building browser extensions for education, I'd love to hear what's worked for you in the comments.