Workflow & routines
Suparanku MCP: let your AI assistant pull your… | Suparanku
Suparanku's MCP server lets any AI assistant — ChatGPT, Claude or Gemini — securely read your AI-visibility data through one endpoint. Connect once with a token, then just ask it to pull this week's recommendations and metrics. It works on every paid plan, inheriting your exact dashboard permissions.
What MCP is, in one sentence
MCP (Model Context Protocol) is an open standard — introduced by Anthropic in late 2024 and now supported across ChatGPT, Claude, Gemini and other clients — that lets an AI assistant securely connect to an outside service and use its data and tools. Think of it as a USB-C port for AI: one socket, and any compatible assistant can plug in.
Suparanku now exposes exactly that socket. Your AI-visibility data — visibility, position, sentiment, competitors, sources, and the recommendations our Growth engine produces — is available to your AI assistant through a single secure endpoint.
Why this is genuinely useful, not just a buzzword
The data is where you already work. Most marketers don’t want one more dashboard to remember to open. With MCP, you ask the assistant you already use — “what changed in our AI visibility this week?” — and the numbers come back in the same chat, in plain language, without a tab switch or a CSV export.
And it matters that this is easy to repeat, because AI answers are not stable. Independent research by SparkToro (Rand Fishkin, ~3,000 runs across ChatGPT, Claude and Google AI) found there is less than a 1-in-100 chance an AI returns the same list of brands twice for the same question, and less than 1-in-1,000 in the same order. A single check tells you almost nothing; what matters is the trend across many runs over time. Checking that trend manually every week is a chore — so most people skip it. When pulling it is a one-line request to an assistant, “every week” actually happens.
A simple example (works on Starter)
Meet a mid-market marketing lead — call her Tanaka-san. Her 社長 has started asking, “are we showing up in ChatGPT?” She’s on Suparanku’s Starter plan, which includes full MCP access — read and write. Her weekly digest only needs to read, so she’ll use a read-only token. Here’s her week:
- One-time setup (2 minutes). In Settings → MCP & API tokens she creates a token and pastes it into her AI assistant (Claude Desktop, ChatGPT, or any MCP client). Hosted clients like claude.ai connect via OAuth instead — no token to copy. Done once.
- Every Monday morning, one message: “From Suparanku, pull this week’s visibility changes for our brand and the top 3 recommendations, then summarize it in three lines for my manager.”
- The assistant does the rest. It calls Suparanku’s read tools, gets the latest metrics and the recommendations our Growth engine already prepared, and writes the summary — which she forwards to her 社長, or drops into Slack.
No dashboard login, no copy-paste, no spreadsheet. The recommendations were generated by Suparanku; MCP just delivers them into the tool she already lives in. Want it fully hands-off? Clients like ChatGPT support scheduled tasks, so the same Monday-morning prompt can run on its own and post the digest before she’s at her desk.
That’s the Starter-tier value in one line: the analysis you’re already paying for, pulled automatically into your assistant, every week, by just asking.
What you can pull — and what needs write access
With a read-only token, your assistant can:
- Read the 5-axis analytics (visibility, position, sentiment, competitors, sources) in any cut.
- Read the recommendations Suparanku has generated for each topic.
- Browse the cited sources AI is using, and your past-scan history.
Write access is included on every paid plan too — mint a token with write scope (or grant it during the OAuth consent), and the assistant can also act: register brand spelling variants, add prompts and competitors, generate new content briefs, and run manual scans on demand. Same channel, more verbs.
Connecting is safe by design
The MCP channel does not invent a new permission model — it inherits the one you already have:
- Same permissions as the dashboard. Row-level security and your role (Owner / Admin / Member / Viewer) are re-checked on every single call. A read-only token cannot write; a Viewer cannot do what a Viewer can’t do in the UI.
- You stay in control. An owner or admin can revoke any token’s or connected app’s access to their company at any time, from Settings → History & Access.
- Two ways in, one channel. Paste-a-token clients use a Personal Access Token; hosted clients (claude.ai, ChatGPT) connect via OAuth 2.1. Both resolve to the same data and the same limits.
Try it
If you’re already on Suparanku, open Settings → MCP & API tokens, create a token, and ask your assistant for this week’s recommendations. Even on Starter, the weekly digest is one message away. For agents and developers, the always-current tool reference lives at app.suparanku.com/docs/mcp and the endpoint is app.suparanku.com/api/mcp/v1.