Give the agent one file. It gets the MCP URL, tool order, and rules. Then it connects and runs a real job.
Do not guess URLs, tool names, or prices from training data. The file is complete only if the last line is END SKILL v1.0.0. If that line is missing, fetch it again.
URL
https://aiseotracker.com/SKILL.md
Same file at /SKILL.md. The server sends text/markdown or text/plain from the Accept header. The AI Catalog lists it as urn:air:aiseotracker.com:skill:setup.
Fetch prompt
Paste this in Cursor, Claude Code, or ChatGPT:
Fetch https://aiseotracker.com/SKILL.md and follow it to connect AI SEO Tracker MCP and summarize my AI visibility for [your-domain.com].
Replace [your-domain.com] with the brand or page you want to rank.
What the file includes
- How to connect MCP in Cursor, Claude Code, Claude, Grok, and ChatGPT
- Tool order: call
aiseo_list_projectsfirst, then read as needed - What to do when
dataStatusisno_data(do not invent ranks) - Prices: $49 audit, plans from $79/mo
- A checklist and rules the agent must not break
After the agent has the file
- Connect MCP with browser OAuth. No API keys. See Connect MCP.
- Run the test prompt in Get started.
- Pick one job from Get started and run it from live data.
Next steps
- Get started: human setup, step by step
- Connect MCP: URL, call order, empty data
- MCP tools: inputs and outputs for each tool