Connecting the MCP

Q: What can my AI assistant do over the Rubric MCP?

It reads your Rubric results (page scores, pillar breakdowns and open issues) and it can act on them: start a crawl, re-crawl a site after a fix, and lint a draft before you publish. So your assistant can run the whole loop, audit, fix, re-crawl and compare, from inside the tool you already use. What it cannot do is change your account: it never edits your settings or billing, never deletes or shares anything, and only ever reads your own data.

Q: Where do I get my MCP API key?

From the MCP page in Rubric. Generate a key there, then add it to your MCP-compatible AI client along with the hosted MCP endpoint. Treat the key like a password: anyone who holds it can read your Rubric results and start crawls on your account, so store it securely and do not share it.

Q: Can the MCP change my account or billing?

No. The MCP can read your results and start crawls or draft checks, but it cannot change your settings, your plan or your billing, and it cannot delete or share anything. Those stay inside Rubric, done by you.

Connect the hosted Rubric MCP so your own AI assistant can read your results and run the fix loop.

By Chris McCarron·5 min read·Updated 5 Oct 2026·Applies to Free and Pro

The Rubric MCP is a hosted service that lets your own AI assistant read your Rubric results and act on them. You connect it with an API key generated on the MCP page in Rubric, then point an MCP-compatible AI client at the hosted MCP endpoint. Once connected, your assistant can pull your scores, pillars and issues, and it can start a crawl or lint a draft on your behalf, so you can run the whole fix loop without leaving the assistant.

In short

→The hosted Rubric MCP lets your AI assistant read your Rubric results and act on them: it can start a crawl, re-crawl after a fix, and check a draft.
→You connect it using an API key generated on the MCP page in Rubric.
→It cannot change your account, settings or billing, and it only ever reads your own data.

What is the Rubric MCP?

MCP (Model Context Protocol) is an open standard that lets an AI assistant connect to an external tool or data source. The Rubric MCP is a hosted MCP built on that standard. It gives an MCP-compatible AI client a way to read your Rubric results directly and to run crawls and draft checks for you, so you can work your citability from inside the assistant you already use, rather than copying numbers out of the report by hand.

The host never scores your site: the Rubric engine computes every number deterministically, and the MCP returns what the engine produced. Your assistant reads those results and can ask Rubric to run a new crawl, but it is never the thing doing the scoring.

How to connect the MCP

Connecting comes down to generating a key and giving it to your AI client.

Open the MCP page in Rubric. This is where your MCP access is managed.

Generate an API key. The key authenticates your assistant as you, so Rubric knows which results to return and whose crawls to run.

Add the hosted MCP endpoint and your key to your AI client. In any MCP-compatible client, add the hosted MCP endpoint and paste your API key so the client can connect.

Note

Steps inside a specific AI client vary by client. Add the hosted MCP endpoint and your API key using that client's own way of adding an MCP server. The details you need, the endpoint and your key, both come from the MCP page in Rubric.

Before you connect

Your API key is a credential. Anyone who holds it can read your Rubric results and start crawls on your account, so store it securely, do not paste it into shared documents, and generate a fresh key rather than reusing one you cannot account for.

What your assistant can do

Once connected, an AI assistant can read the same results Rubric shows you: your page scores against the 70 line, the pillar breakdown across Known, Findable and Trusted, and the open issues behind each check. It can also act on them: start a crawl, re-crawl a site after you ship a fix, and lint a draft before you publish. That means you can ask your assistant to run the whole loop, find your weakest pillar, make the top fix, re-crawl, and compare the before and after, and it works from your real Rubric data.

The MCP returns results; it does not recompute them. The scores it shows are from the latest crawl of that site, so ask it to re-run a crawl (or run one in Rubric) when you want fresh numbers.

What the MCP cannot do

The MCP can read your results and run crawls and draft checks, but it cannot change your account: it cannot edit a setting, change your plan or billing, delete your data, or create or change a share link. Every account change stays inside Rubric, done by you. And an API key only ever reaches your own data, never anyone else's.

MCP, Copy for AI and CSV: when to use which

Rubric gives you three ways to take your results elsewhere, and they suit different needs.

Method	Best for
MCP	Letting your own AI assistant read your live Rubric results and run the fix loop on demand
Copy for AI	Copying your money queries as text to paste into an AI tool for a quick test
CSV export	Pulling scores and failing checks into a spreadsheet or your own tracker

For the text-paste route, use Copy for AI in the report. For spreadsheets, export to CSV and Sheets. The related articles cover both, and what Rubric checks explains what the results themselves mean.

Common questions

What can my AI assistant do over the Rubric MCP?+

Where do I get my MCP API key?+

Can the MCP change my account or billing?+

Did this answer your question?

Exporting to CSV and Sheets

Using Copy for AI

What Rubric checks