AI Visibility MCP for n8n

In n8n, the raw answers the six engines actually give become a workflow step. Wire the MCP Client Tool node into an AI Agent and it can call fetch_raw_answers; the answer text and every cited source come back as structured data your later nodes can diff, store or route. One MCP server, one deliberately narrow tool — and no browser to babysit on a schedule.

Connect AgentGEO to n8n

In an AI Agent workflow, add an MCP Client Tool node and point it at the AgentGEO server over the command-line (stdio) transport — command npx, arguments -y agentgeo-mcp --key ag_live_...:

{
  "command": "npx",
  "args": ["-y", "agentgeo-mcp", "--key", "ag_live_..."]
}

The stdio transport needs a runtime that can execute npx, so this path suits self-hosted n8n. Create your ag_live_ key in the quickstart — free tier, no credit card. Building the workflow first? An ag_test_ key always returns clearly labelled demo records at zero credits, so you can wire up and test the nodes before spending anything.

Either path returns one record per surface — the provider's answer, verbatim, with every source it cited as structured data (the full record also carries the raw providerFields):

{
  "surfaceKey": "perplexity",
  "status": "delivered",
  "answerText": "...",
  "sources": [
    { "title": "...", "url": "https://...", "position": 1 }
  ],
  "fetchedAt": "2026-08-19T09:12:04Z",
  "latencyMs": 8412,
  "providerRecordId": "..."
}

What to build once it's connected

This is where n8n shines: the check runs itself. A few workflow recipes to start from (build them on an ag_test_ key, then swap to ag_live_ to go live):

• Weekly six-engine sweep to Slack. Schedule Trigger → fetch your money queries across all six surfaces → format a citation table → Slack node. A standing visibility report with no dashboard to log into.
• Competitor-domain watch to a Sheet. Cron → fetch your category query → filter sources[] for competitor hosts → append a dated row to Google Sheets. A longitudinal record that's yours.
• Drop-out alert. Daily fetch of Perplexity for your main query → check whether ourdomain.com is still in sources[] → IF it isn't, send an email or Slack alert. Know the day you fall out of the answer.
• AI Mode vs Overviews diff. Fetch google_ai_overview and google_ai_mode for the same query → diff the cited URLs → post what AI Mode adds.
• New-content trigger. After you publish, fetch the target query across engines a week later → append the citation result so you can see whether the page earned its way in.
• Agent-summarized digest. Feed the fetched records to an AI Agent node and have it write a plain-English 'what changed this week' summary before the Slack step.

What the MCP server gives you

Deliberately thin — an answer-access layer for your workflow, not an analytics product bolted onto n8n.

• One narrow tool — fetch_raw_answers posts to AgentGEO's API and returns the run envelope verbatim. Nothing else to learn, nothing else to break.
• Six engines behind one contract — ChatGPT, Perplexity, Gemini, Google AI Overviews, Google AI Mode and Copilot with an identical record shape, so widening a sweep is one more value in the surfaces array.
• Structured `sources[]` — every citation with its title, URL and position, ready for downstream nodes to filter, diff or write to a Sheet. No links regexed out of prose.
• Raw records only — no rankings, sentiment or visibility scores. The conclusions come from your workflow — a Function node, an AI Agent step, or your own logic.
• Managed-scraper engine, maintained for you — under the hood the server talks to POST https://api.agentgeo.org/v1/fetches; keeping six collection paths healthy is AgentGEO's job, not a headless browser in your automation.
• Usage-based billing with a spend cap — one credit per delivered record, failed records cost zero, free tier with no credit card, never per-seat.

What it doesn't do

The narrowness is the point. Your workflow holds the logic and destinations; conclusions computed outside that context would be worth less than raw records you route yourself — that is the whole GEO data layer argument. So, honestly:

• No dashboards, no scores, no recommendations engine. The records are the input; the report, the alert and the Sheet row are what your workflow builds from them.
• Not instant — set generous node timeouts. Requests wait up to 180 seconds because live surfaces are slow: an AI Overview SERP round-trip runs 40–90s. Give the MCP or HTTP Request node enough timeout to let a multi-surface fetch finish.
• Some fetches finish async. A slow chatbot scrape can exceed the sync budget and come back failed with a providerFields.snapshot_id; a follow-up call with that id and the same single surface redeems the finished answer without paying again (not valid for google_ai_overview) — worth a retry branch in a monitoring workflow.

Prefer the REST API?

The HTTP Request path is the REST API: POST https://api.agentgeo.org/v1/fetches, the same endpoint the MCP server wraps. The curl guide shows the exact request, and the Python guide helps if a Code node does the heavy lifting — identical contract either way, so a Cloud workflow and a self-hosted one behave the same.

Who builds on this

Automation builders and marketing-ops folks who want GEO monitoring as a workflow, not a codebase. If your instinct for 'track how the six engines answer for our category' is a Cron trigger and a Slack node rather than a repo, n8n plus AgentGEO gets you a standing monitor in an afternoon. Try a single fetch first in the no-signup playground, or see how the approaches compare.