Your visibility numbers, your gap list and your published evidence, reachable in a conversation — across every client in one sentence, combined with whatever else you tell Claude about the business, done in one instruction instead of five dashboard screens. Eleven read tools are on from the moment you connect. Five more can act — correct the entity, close a gap, approve a publish, start a run — once you tick the box that starts unticked. Every write is capped at 30 an hour and logged where you can read it.
An agent key or signing in — both on every plan, Starter included, and in the trial.
Not because it knows a different number — every number it reads is one the dashboard already shows. Because of what changes when those numbers are reachable through a conversation.
One sentence, every project
Every read tool defaults to the whole account and narrows to one project only when you say so — an agency’s whole book in one question, not a project switcher clicked twelve times.
It knows what you tell it, too
A dashboard cannot read the board deck open in your other tab. Claude can hold that and your visibility numbers in the same conversation, and write the sentence that connects them.
Routines, not screens
Read the gaps, decide, close one, list what waits, approve the one you’ve read — five dashboard screens become one instruction, because the five write tools exist to be asked for.
Built for agency scale
A connection bound to one client’s project cannot see another, even if asked — enforced at the one door every tool goes through, not by convention.
And the honesty travels with the number: every rate still carries its sample size, because a rate over three prompts is not a measurement — stated to the calling model itself, not only in a help article it could skip.
The exact prompt, the tools it calls, and what comes back. Copy one and try it once you're connected.
“Give me a one-paragraph Monday update across every client project: who's visible, who's slipping, and what's waiting on me.”
get_visibility_summaryget_gapslist_waitingOne paragraph spanning every project in the account, each rate with its sample size, closing with what is waiting for a decision.
“For example.com, which topics are we losing and to whom? Close the top one.”
get_topic_visibilitylist_cited_sourcesget_gapsclose_gapNames the topic and the site cited instead with its sample size, then asks the Evidence Engine to work the highest-impact gap now.
“We just launched a fixed-fee onboarding package called 'Fast Start' — a one-week, done-for-you setup, $500 flat. Update our entity with that and confirm the Fast Start product.”
get_entityconfirm_productupdate_entityRecords the product from your own words, updates the entity, and queues the canonical page on your own site to be rewritten from it.
“Show me what's waiting for approval on example.com. I've already read the Fast Start rewrite in the app — approve that one, leave the rest.”
list_waitingapprove_publishSends only the one item you named to your live site. Everything else on the list stays exactly as waiting.
“Run a fresh visibility check on example.com and tell me what changed since last time, with sample sizes.”
run_visibility_nowget_visibility_summaryStarts a run on your own AI keys, then compares it with the last one, engine by engine — and says when a move is too small a sample to call a trend.
“What's our visibility on client.com this week, and is there anything waiting for me?”
get_visibility_summarylist_waitingConnect on a client’s behalf and choose ONE project at consent: every tool then answers for that client alone, and naming a different client is refused.
“Across all my topics, which sites are citing us the most, and which only ever cite our competitors?”
list_cited_sourcesA ranked source list with each site’s kind and its naming rate — a stockist and a competitor are never folded into one figure.
“Here's our board deck draft for Thursday [pastes the text]. Pull in this month's visibility numbers and gap closures and write three bullets for the AI-visibility slide.”
get_visibility_summaryget_published_evidenceThree board-ready bullets in the deck’s own voice, each stat still carrying its sample size — something no dashboard alone can write.
Two ways in, both on every plan including Starter and in the trial, both scoped to your account alone.
An agent key
For any MCP client, including ones that can't sign in. Settings → Agent access → create a token, choose what it may read, leave or tick “Let it make changes” (unticked by default), and copy it — shown once, only its hash is kept.
Sign in (OAuth)
For a client that supports MCP sign-in — Claude does. Add https://therankingfactory.com/api/mcp as a custom connector; there is no token to copy. Choose every project or one, and leave the write tick unticked unless you want it.
Connecting on a client's behalf? Choose one project at consent instead of every project — the connection then reads (and, if you tick it, writes) that one client and nothing else, refused out loud if it's ever asked about another.
Is The Ranking Factory's MCP server read-only?
No. It reads everything in the account from the moment a credential connects, and can also write once you opt that connection into a separate scope — a box that starts unticked on the sign-in consent screen and when you mint a key.
Do I need a specific plan to connect an AI assistant?
No. An agent key and signing in are both available on every plan, Starter included, and during the 14-day trial. Choosing one project or every project happens at connection time either way.
What can an agent actually change?
Only what the write scope unlocks, and only once you tick it: update_entity, confirm_product, close_gap, approve_publish and run_visibility_now — five tools, each calling the same code your own dashboard buttons call, with the same refusals.
Is every write logged?
Yes. Every write attempt — done, refused or failed — is recorded with the credential, the project, the tool and the outcome, and shown to you under Settings → Agent access → What agents changed. Writes are also capped at 30 an hour per credential.
Connect Claude in a couple of minutes, on any plan. Read from the first question; write only once you say so.
Start your 14-day free trial