Use case · Content & copywriting
Connect Claude, ChatGPT or your content brief tool to Grades.com through the MCP server or the REST API. Your agent reads the business's reviews and turns them into theme briefs, quotes with their source, FAQ answers and refresh checks. Every quote it returns is a real review you can trace.
review_insightsRepair reviews praise: photos of the damage, a quote the same day.review_showcase"They sent photos of every cracked shingle before quoting." Google, Sep 18.review_searchFour reviews mention a tarp put up the night of the storm.Pack: 3 quotes, 2 paraphrases, the 6 reviews behind them
Copy that sounds like customersThe words people use for a service, not the words a model guesses for the category.
Evidence with a sourceEvery quote comes back with its site and date, so an editor can check it in one click.
Read-only by defaultA writing agent needs a key that reads reviews, nothing more.
What your agent can do here
Each one is a sentence typed into Claude, ChatGPT or your brief tool. The agent calls the Grades.com tools named under it and answers from the business's reviews.
"From the last 12 months of reviews, what do customers praise, in their own words? Rank it by how often, and tell me which services get the most specific praise." review_insights returns the top praise and complaint themes for the period; review_tags_list counts the reviews per service tag. Write the page around what people already value.
"Three direct quotes with site and date for storm repairs, plus two indirect points I can paraphrase without quoting." review_showcase picks reviews by rules and never edits them; review_search finds the ones behind each paraphrase, so your editor can check both.
"What do customers ask or complain about again and again? Turn the top four into FAQ answers." The complaint themes from review_insights and phrase searches with review_search show what people wanted to know before they bought, and what surprised them after.
"Our gutter page promises next-day installs. Do recent reviews still back that up?" Your agent reads the page itself, then checks each claim against recent reviews with review_search and the monthly trend from review_stats. Claims the reviews no longer back up come back with a suggested rewrite.
"Rewrite the intro so it shows first-hand experience: one quote, one paraphrase, and list the reviews you used." The agent works like an agency writer with a folder of reviews, through review_showcase and review_list, and hands back the sources with the draft.
Connect it once
Add the MCP server to Claude, ChatGPT or Cursor and sign in, or give your brief tool a read-only API key. Webhooks can tell your pipeline when a new review lands. Writers without an AI client can use Ask Grades, the assistant in your dashboard, for the same questions.
MCP server https://mcp.grades.com/mcp
REST API https://app.grades.com/api/v1
Scopes reviews:read is enough for every workflow here
ChatGPT custom connectors need a Business, Enterprise or Edu plan
Using reviews in copy
A quote in quotation marks, word for word, with the site and the date. Short is better: one sentence that proves one point.
The themes behind the copy: what customers keep praising, in plain words, without quoting anyone. Keep the reviews behind a paraphrase so an editor can check it.
Review sites set their own rules for reusing what is posted there. Check the site's terms before you publish a quote. Our review-site policies page is a starting point, not legal advice.
The agent gets real reviews with their source and nothing else. Review Showcase rules never edit, shorten or merge a review and never return Yelp text.
First-hand experience
Google's guidance on helpful content asks for first-hand experience. Real customer details, with a source, are one honest way to show it. We make no claim about what it does to rankings.
For a healthcare business, conditions, treatments and amounts of money are replaced with placeholders before review text reaches any model Grades.com runs, and Review Showcase never offers a review that names a treatment as a quote.
Review Radar finds review pages you do not track yet, and the Review Archive keeps each review's earlier versions, so you know what a quote said when you used it.
What we can and cannot do. Better copy is yours to write; nobody controls what a search engine or an assistant does with it, and we never promise to change an answer.
What the content team gets
Each one runs per location, in the dashboard and through the MCP server and the API.
| Deliverable | Product | MCP tools | Plans |
|---|---|---|---|
| Praise and complaint themes, service tags and sentiment for any period | Insights | review_insights, review_tags_list | Every plan: Google on Starter, every site from Pro |
| Every review, searchable, with its earlier versions | Review Archive | review_list, review_search, review_history | Every plan |
| Review pages you do not track yet, so the evidence covers every site | Review Radar | radar_mentions | Weekly search on Pro and Premium |
| The same questions in the dashboard, for writers without an AI client | Ask Grades | None to set up | 50 questions per location a month on Starter, 300 on Pro, 2,000 on Premium |
| Your own agent or brief tool, connected, with quotes picked by rules | Connect your AI | review_showcase, review_stats | Every plan |
Not from our side. The tools return real reviews with their site and date, and a quote picked by rules is never edited. A model can still misquote what it read, so ask it to list its sources and have an editor check each quote against the review before it goes live.
It depends on the site the review was posted on. Each review site has its own terms on reuse, so check the site's terms before you publish a quote. Paraphrasing what many customers say, without quoting anyone, is the safer default.
Grades.com replaces conditions, treatments and amounts with placeholders before review text reaches a model we run, and Review Showcase never offers a review naming a treatment. Your own agent can read the full text through the API, so keep patient details out of anything you publish.
It gives your writers first-hand material, which is what that guidance asks for. Whether a page ranks depends on far more than this, and we make no promise about it.
Product names and logos are trademarks of their owners; no endorsement implied.