Automated SEO competitor analysis does not need a dashboard. If you already work in Claude Code, Cursor or Codex, your agent can call the lookups, collect the results and draft the action list while you review it. This guide sets that up with seogeoaeo.ai's MCP server, puts a spending cap on every call, and maps each step of the analysis to a tool slug.
Everything here is from seogeoaeo.ai's tool reference and pricing page, checked in October 2026.
Why an agent fits this job #
A competitor analysis is a fixed sequence with a judgment call at the end. Find the rivals, pull their keywords and top pages, pull their links, check technical basics on a few pages, check AI citations, then decide. The first five steps are lookups. An agent can run them in order, keep the outputs in your repo, and leave you the part that needs a person: choosing three or four actions.
Connect the MCP server #
seogeoaeo.ai exposes every tool over MCP at https://seogeoaeo.ai/api/mcp and over REST at https://seogeoaeo.ai/api/v1. Create an API key in your workspace (keys start with sga_ and are shown once), then add the server. For Claude Code:
claude mcp add --transport http --scope user seogeoaeo https://seogeoaeo.ai/api/mcp --header "Authorization: Bearer YOUR_API_KEY"
Cursor and Windsurf read the same server from an mcpServers entry, VS Code from .vscode/mcp.json, and Codex from ~/.codex/config.toml with the key in an environment variable. Claude Desktop connects through mcp-remote.
Each tool becomes an MCP tool named by its slug. Agent runs cost the same as runs in the browser.
Put a cap on spending #
Two arguments make automated runs safe to leave alone:
maxCreditsrefuses a run above the price you set.idempotencyKeylets the agent retry a call without paying twice.
The check-credits tool is free, so tell the agent to read the balance first, add up the planned runs and stop if the total is above your limit. A run that fails before producing a result is refunded.
The steps, as tool calls #
| Step | Tool slug | Credits | What comes back |
|---|---|---|---|
| Find search rivals | competitor-finder |
38 | 30 sites sharing the most Google keywords with your domain, with overlap share |
| Rival keywords and top pages | competitor-keywords |
35 | Top 100 keywords by estimated traffic, positions, intent, ranking pages |
| Rival size at a glance | domain-overview |
50 | Keyword count by position, estimated visits, linking sites, link rank |
| Rival links | backlink-finder |
70 | 100 strongest linking sites with spam score and first-seen date |
| Page basics | on-page-seo-checker |
8 | Title, description, snippet clipping, heading outline |
| Content without JavaScript | ai-crawler-view |
8 | Raw HTML compared with the rendered page |
| Mobile speed | core-web-vitals-snapshot |
20 | Field LCP, INP and CLS when available, plus a Lighthouse lab run |
| AI citations | ai-citation-finder |
160 | Pages Google AI Overviews and ChatGPT cite most for a topic |
Every succeeded run also returns improvementTips, a list ordered by priority with the change to make and the URLs it affects. For your own pages, that is a head start on the fix list.
A prompt that works #
Give the agent the order and the stopping rule in plain words:
Check credits. Run competitor-finder on example.com for the US. Take the four closest competitors, skipping very large general sites. For each, run competitor-keywords and backlink-finder with maxCredits set to the listed price. Run on-page-seo-checker and ai-crawler-view on their top page and on our closest page. Run ai-citation-finder on our main topic. Save every result as JSON in research/competitors/. Then draft three or four actions, each with the evidence, and stop.
For four rivals that plan costs 38 + 140 + 280 + 160 = 618 credits, plus 8 credits for each page check. The Indie plan gives 2,000 credits for $29 a month.
Keep the judgment human #
The agent will happily produce twenty actions. Cut them to three or four. Look for gaps that repeat across rivals, positions you can realistically beat (ninth place, not second), and link sources you could also earn. Treat every volume and traffic figure as an estimate, useful for comparing rivals but not as a count.
Limits to plan around #
seogeoaeo.ai does not track rankings over time and Competitor Keywords cannot read Search Console data. The AI Citation Finder covers Google AI Overviews and ChatGPT, not Perplexity or Gemini; ai-answer-visibility samples ChatGPT, Perplexity and Gemini for one question at 120 credits. If you want daily rank history for your rivals, a rank tracker such as Semrush Position Tracking is the better fit for that part.
Rerun the whole script each quarter and diff the JSON. That diff is the most useful page in the folder, and it is the cadence seogeoaeo.ai's own step-by-step guide to the process recommends.