AI SEO Tracker measures how ChatGPT, Perplexity, Gemini, Copilot, and AI Overviews mention a brand. From Cursor, Claude Code, Grok, or Codex you can read that same data without opening the dashboard.
You work with it like a teammate: connect MCP, give it a job with a clear result, review the draft.
MCP is the live tool connection (OAuth, no API keys). Tools are namespaced
aiseo_*. The setup file is/SKILL.md. The only write isaiseo_add_prompts. MCP does not send email, add competitors, or charge a card.
MCP is account-scoped
One OAuth session covers every project that user can open. You copy the URL from a project’s Agent settings, but the token is not limited to that project.
Treat a connected agent like a login on a shared computer: revoke it at /account/agents if you do not want it to keep reading.
The projectSlug you pass to tools is how you pick which workspace to read. Call aiseo_list_projects first. Never invent a slug from the company name.
A good first handoff
Start with a real task that has a finished deliverable:
Call aiseo_list_projects. Use the first project. List missing prompts and the top citation domains. Summarize what to do this week. Do not invent ranks.
This tells the agent what to do, where to read, and when to stop.
If you want a repeatable format, pick a job from Get started.
What MCP cannot do
- Add competitors or change settings
- Send email or post to Reddit
- Charge a card
- Invent ranks, URLs, or rival names when
dataStatusisno_data
aiseo_add_prompts can add queries the user asked to track. That may start scans if the account is entitled. It still cannot invent prompts.
Empty JSON with dataStatus: "no_data" means the project has no scan data yet. That is not a failed connection. Call aiseo_get_scan instead of inventing ranks. done means no in-flight jobs, not that every prompt has results.