MCP connector

Your SEO data, inside your AI assistant.

Connect Rookery to ChatGPT, Codex, Claude, or another compatible MCP client. Thirteen focused tools let your assistant inspect crawl evidence, start crawls, draft fixes, create tasks, and prepare reports. Any live-site deploy still requires your explicit approval inside Rookery.

https://rookeryseo.com/api/mcp

One URL for every assistant. You sign in with your normal Rookery account — no API keys.

Paste the URL

Add the connector URL wherever your assistant asks for a connector or MCP server. Step-by-step guides below.

Sign in with Rookery

A normal sign-in screen opens — your Rookery email and password. The assistant only sees the sites your account can see.

Ask about your sites

“What did the last crawl find?” “Draft fixes for the missing titles.” Rookery returns current evidence and app links for review.

Setup

Pick your assistant.

Claudeclaude.ai on the web and the Claude Desktop app — same steps.

1
Open the prefilled Rookery connector or go to Claude's Settings, then Connectors.
2
If you opened Settings, click Add custom connector and paste the connector URL above.
3
Click Add, then sign in with your Rookery account when prompted — no API keys or setup files.

Claude CodeOne terminal command.

In your terminal, run:

claude mcp add --transport http --scope user rookery https://rookeryseo.com/api/mcp

2
Start Claude Code, type /mcp, and choose rookery to authenticate.
3
Sign in with your Rookery account when prompted.

ChatGPTCustom MCP connectors need developer mode turned on once.

1
In ChatGPT, open Settings, then Connectors.
2
Open Advanced settings and turn on Developer mode (a one-time toggle).
3
Back in Connectors, click Create, name it Rookery, and paste the connector URL above as the MCP server URL.
4
Choose OAuth for authentication, click Create, then sign in with your Rookery account when prompted.

CodexOne command for the Codex app, CLI, and IDE extension.

In your terminal, run:

codex mcp add rookery --url https://rookeryseo.com/api/mcp

2
Run codex mcp login rookery if Codex does not open the Rookery sign-in flow automatically.
3
Finish signing in with your Rookery account in the browser.

CursorAdd one snippet to your mcp.json.

1
Open Cursor Settings, then Tools & Integrations, and click New MCP Server.

Add this to the mcp.json file Cursor opens:

{
  "mcpServers": {
    "rookery": {
      "url": "https://rookeryseo.com/api/mcp"
    }
  }
}

3
Save the file, then click the sign-in prompt next to rookery in the MCP list.
4
Sign in with your Rookery account when prompted.

A custom install makes Rookery available to your assistant. It does not make Rookery globally discoverable or guarantee that an assistant will suggest it; public discovery depends on OpenAI and Anthropic directory review and placement.

Capabilities

What it can do.

Rookery exposes 13 focused MCP tools: eight read-only discovery, evidence, status, and export tools, plus five actions for crawling, drafting, task creation, deploy proposals, and report generation.

See your sites and recent audits
“Which of my sites has the most open issues?”
Summarize an audit
Scores, top issues, and quick wins from the latest crawl.
Dig into pages, issues, and links
Page-level evidence, the full issue list ranked by impact, and the internal link graph.
Pull themed reports
On-page, indexability, links, and performance — already aggregated.
Start a crawl
“Re-crawl acme.com and tell me what changed.” Your plan’s audit limits apply.
Draft fixesPro · uses AI budget
New titles, meta descriptions, and H1s land in your Fix Queue for review.
Add issues to your Fix Queue
Saves selected issues as tasks in Rookery. It does not change the live site.
Propose a live-site deploy
Creates a seven-day review card. Only you can approve it inside Rookery; the MCP has no approval or deploy action.
Generate a client-ready reportPro · uses AI budget
Creates or refreshes the stored AI audit report and returns a share link.
Export CSV evidence
Returns a short-lived signed download link for pages, issues, or links.

Honest limits: reading audit data and CSV export work on every plan. AI drafting through an assistant draws from the same AI budget as drafting inside the app, and drafting and AI reports need a Pro plan — the same rules as everywhere else in Rookery.

It can act. It can never approve its own deploy.

Your assistant signs in with your own Rookery account and sees only the sites that account can see. It can start crawls, create tasks, draft fixes, and generate reports when you ask; those actions can use crawl quota or AI budget. A live-site change is different: the connector can only create a temporary deploy proposal. The user must review and explicitly approve it inside Rookery, and the MCP has no tool that can approve or execute that proposal.

You can stop access by removing the connector inside the assistant. For detailed uninstall, server-side revocation, privacy, and incident-reporting steps, see MCP support and security.

FAQ

Straight answers.

Is it safe to connect my account?

The connector uses OAuth — you sign in with your Rookery account on a Rookery page, and the assistant never sees your password. It can read your crawl data and propose changes, but nothing is published to your site without your approval inside Rookery. You can revoke access anytime by removing the connector in the assistant or signing out sessions in your account settings.

Which plans include the connector?

Every plan can connect, read audit data, and export CSVs. AI drafting through an assistant uses your plan's AI budget, and drafting and AI reports need Pro — the same rules as inside the app. No separate add-on to buy.

Does it work with assistants other than Claude?

The documented clients are ChatGPT developer-mode plugins, Codex, Claude on the web and desktop, Claude Code, and Cursor. Other clients may work if they support remote Streamable HTTP MCP servers and OAuth metadata discovery, but their current release should be tested before relying on it.

Do I need an API key?

No. When the assistant first uses the connector, a Rookery sign-in screen opens. Sign in once and you're done — nothing to generate, paste, or rotate.

Can the assistant change my website?

It can take requested account actions such as starting a crawl, creating a draft or task, and generating a report; those may use plan quota or AI budget. It cannot approve or execute a live-site deploy. The deploy tool only creates a temporary proposal, and publishing happens only after you explicitly approve it inside Rookery.

What data does the assistant see?

Exactly what your Rookery account sees: your sites, audits, pages, issues, links, reports, and the records created by actions you request. It cannot see other Rookery customers, your payment card details, or anything on your computer. Your assistant provider handles retrieved information under that provider's own account settings and privacy terms.

Will a custom install make Rookery appear for every user?

No. A custom connection makes Rookery available only to that user or workspace, and the assistant still decides whether a relevant request needs it. Public discovery requires separate OpenAI plugin approval or acceptance into Anthropic's Connectors Directory. Directory placement and proactive suggestions are controlled by each platform and are not guaranteed.

How do I remove access or report a security issue?

Remove Rookery from the assistant to stop that client from using the connection. Server-side revocation, client-specific uninstall commands, privacy details, and security reporting are available on the public MCP support page.

Two minutes of setup. Every audit, on call.

In the app this lives under Settings, in the AI Assistants tab.