Ask an assistant about your ranks

You can ask an assistant about keyword positions without building a private API client. RankMySEO exposes an MCP endpoint at https://api.rankmyseo.com/mcp. The lobby starts without an Authorization header. You call register, save the bearer token, then send Authorization: Bearer on later requests. If a human already has an account, they mint a lasting key on the Account screen instead. There is no public REST signup for agents to create billed accounts. Do not invent one.

A rank without a stored trail is useless for setting a goal or seeing a trend. MCP does not change that rule. It changes how you fetch the series, the project tools your plan allows, and whether you are on a walk-up guest key or a lasting human account.

Two ways in, not one

Walk-up register. Connect to the MCP URL with no auth. Call register. Store the token. Add a project, add keywords, ask about positions. Walk-up stays limited: one website, two keywords, short guest lifetime, research that does not linger. One guest bot per network. To keep a site on a human plan, use request_link and have them open the URL.

Lasting Account key. A human signs up for Free at the app signup URL, or already has a plan. They open Account, mint a lasting key, and put that token in Authorization. That key uses the account plan. Same gates as the dashboard. Free stays Free. Starter keys unlock Starter tools. Agency and Scale keys can also call track_ai within their prompt limits.

Checkout is open after Free signup. Humans pick a paid plan in the dashboard. Agents do not create billed accounts through a hidden REST path.

Walk-up guest research is short-lived on purpose. The lobby is for trying the MCP tools, not for pretending you already have a billed seat. If an assistant keeps a guest token in a long-running automation, expect the guest limits and deletion windows from the live docs to bite. Move important sites to a human Free account, then mint an Account key for the assistant. That path keeps weekly Google organic history on a lasting login instead of a temporary guest bot.

What you can actually ask

With a lasting key on a capable plan, assistants can ask about keyword positions, rank history, on-page issues, competitors, and backlink review on Starter. Starter tools in the live catalog include keyword gap, competitor traffic, keyword planner, backlink review, rank compare, issue pages, next actions, and research site. Agency and Scale can also use track_ai (10 prompts a site on Agency, 25 on Scale). Disavow stays in the dashboard, not over MCP.

Walk-up guest keys still see tool names, but gated tools tell you to call request_link. Call account_status when you are unsure what the current key can do.

Everything stays Google organic by country and language. MCP is not a Maps client. It is not another search engine. It is not phone ranking. If an assistant invents those surfaces, that is the model hallucinating, not the product expanding.

Cursor and Claude style setup

For Cursor, add the MCP server URL in project or user mcp.json. Reload. Call register. For later sessions, put Authorization: Bearer with your saved token in the server headers.

For Claude Code or Claude Desktop, use a Streamable HTTP server entry pointed at the same MCP URL. Same flow: connect without a key, register, then send the bearer token. Any Streamable HTTP client can POST to the endpoint, list tools, register, then authenticate.

Paste tokens carefully. Do not commit guest tokens into public repos. Prefer lasting Account keys for ongoing human work.

Plan gates beat prompt optimism

Assistants love to promise the full catalog. RankMySEO will not. A Free lasting key is still Free: one site, three keywords, weekly cadence, no card, short history window. Starter tools do not appear because the prompt asked nicely. Agency track_ai does not appear on Starter. Walk-up does not become Free because someone said signup in chat.

When you brief an assistant, say the plan out loud:

Guest walk-up: temporary, two keywords, link a human to keep the site.
Free human account: weekly Google organic, three keywords, lasting login.
Starter and up: competitor and issue tools on the same project.
Agency or Scale: track_ai within listed prompt caps.

That briefing prevents the assistant from writing a fake audit.

What MCP is not

Not a substitute for locale discipline. Country and language still matter.
Not daily Free tracking. Weekly on Free stays weekly.
Not a REST billing API. No agent-side billed signup.
Not a SERP vendor mashup. No third-party rank-engine branding in your answers.
Not permission to invent traffic, ROI, or GSC rows the tools did not return.

If the tool returns a position series, report the series. If history is thin, say a single rank without a trail is useless for a goal or a trend, and wait.

A short operator checklist

Decide guest register versus Account lasting key.
Confirm Google organic locale on the project.
Call account_status before promising tools.
Ask for history, not a single rank, when you narrate movement.
Keep Disavow and checkout in the human dashboard.
One soft product link in human-facing copy is enough.

MCP is a control plane for the same rank data the dashboard shows. It does not replace how-it-works education for humans who still think Free means daily. Keep assistant docs boring and literal. When you want the public MCP landing in one place, use that URL once in the close and leave other product links out of that slot.

Connect the MCP endpoint, register or paste an Account bearer token, and ask about the ranks your plan can actually see: https://rankmyseo.com/mcp

Keep reading

See how RankMySEO works

All posts →