SEO MCP server for AI agents

Give Claude Code, Cursor, Zed, or any Model Context Protocol agent the same 91-rule audit that powers MetricSpot's dashboard. Six tools, one HTTP endpoint, no glue code.

No card needed. Results in 30 seconds.

Or just paste a URL: the same audit, no install required.

Updated

Install the SEO MCP server in your agent

The MetricSpot MCP server is live at https://mcp.metricspot.com over Streamable HTTP, and on npm as @metricspot/mcp-server for local stdio. Most agents take a single config entry. The anonymous audit tool works without any key (1 audit per IP per 24 hours); the other 5 tools accept a Bearer token issued from your MetricSpot dashboard.

Claude Code

Anthropic's official CLI. Restart the session after adding.

claude mcp add metricspot --transport http https://mcp.metricspot.com/mcp

Cursor

Add to ~/.cursor/mcp.json, then reload Cursor.

{
  "mcpServers": {
    "metricspot": {
      "url": "https://mcp.metricspot.com/mcp"
    }
  }
}

Zed

Add to ~/.config/zed/settings.json.

{
  "context_servers": {
    "metricspot": {
      "command": {
        "url": "https://mcp.metricspot.com/mcp"
      }
    }
  }
}

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json on macOS, then restart Claude Desktop.

{
  "mcpServers": {
    "metricspot": {
      "url": "https://mcp.metricspot.com/mcp"
    }
  }
}

ChatGPT

Custom MCP connectors require a ChatGPT Pro, Team, or Enterprise plan. Add under Settings → Connectors.

{
  "name": "MetricSpot",
  "url": "https://mcp.metricspot.com/mcp",
  "headers": {
    "Authorization": "Bearer ms_live_xxx"
  }
}

Gemini

Uses the gemini CLI MCP configuration. Requires the Gemini CLI version that supports remote MCP servers.

gemini mcp add metricspot --transport http https://mcp.metricspot.com/mcp

Six tools exposed by the MCP server

Every tool runs the same audit engine as app.metricspot.com. No feature gap, no second-class data. Tool descriptions, inputs, and outputs are validated by Zod schemas so agents get strict, predictable responses.

run_audit_anonymous No API key

Run a one-shot SEO and AI-readability audit on any public URL. Returns scores across 11 modules and ~90 checks, plus actionable findings with documentation links. Rate-limited to 1 audit per IP per 24 hours.

Input:
url
Output:
Full audit envelope with module scores, findings, and a report URL.
run_audit API key required

Queue a full SEO and AI-readability audit. Includes Core Web Vitals from Google PSI and organic traffic if Google Search Console is linked. Returns immediately with status: queued and an audit_id.

Input:
url
Output:
Audit envelope with audit_id, status, and a poll hint.
get_audit API key required

Fetch a previously-run audit by id. Returns module scores (0-100), the total score, every finding with severity and recommendation text, and links to the HTML and PDF reports.

Input:
audit_id
Output:
Audit envelope with status, total_score, module_scores, findings.
list_audits API key required

List the user's audits, most recent first, deduplicated by URL. Returns audit_id, url, status, total_score, and created_at. Default limit 24, max 100.

Input:
limit (optional)
Output:
Array of audit summaries.
get_audit_pdf API key required

Return a signed download URL for the branded PDF report. If no PDF has been rendered yet, queues a render and returns status: queued so the agent can poll the same tool.

Input:
audit_id
Output:
Signed PDF URL or queued status.
get_organic_traffic API key required

Return the 28-day organic traffic snapshot for an audit if the user has linked GA4 and Google Search Console. Includes session count, daily trend, top landing pages, top queries, and indexing health.

Input:
audit_id
Output:
Organic traffic snapshot or a not-linked status.

Why use an MCP server instead of a REST API or manual audit

MCP is purpose-built for AI agents: tool descriptions, typed schemas, and auth all travel with the connection. The agent can plan a multi-step audit (run, poll, summarize, PDF, follow up on findings) without you writing any glue code.

Feature Manual SEO audit Other SEO APIs MetricSpot MCP
Setup time Hours per site, repeated each run Hours: read docs, write client, handle auth, parse JSON One config line in your MCP client
Agent-native No: human in the loop for every step Partial: agents can call REST, but need hand-written tool wrappers Yes: tool descriptions, schemas, and rate limits travel with the connection
AI-readability checks Rare: most templates predate llms.txt and AI Overviews Limited: most APIs focus on backlinks and keywords 11 dedicated AI rules, scored alongside traditional SEO
White-label PDF Manual export, manual rebrand Usually not offered, or on the highest tier get_audit_pdf returns a branded PDF on every paid plan
Anonymous trial Not applicable Almost never available without a paid account run_audit_anonymous works with no key, 1 per IP per 24 hours
Pricing model Time billed by the hour Per-call metering, often per-keyword or per-domain Flat monthly plans aligned with the dashboard

MCP server pricing

The MCP server is included on every paid MetricSpot plan: same audit credits, same modules, same PDF rendering. The anonymous audit tool is free and works without an account. Authenticated tools count against the same monthly audit quota you see in your dashboard.

Free

$0/mo

Try the platform. No card, no commitment.

  • ·10 audits per month (1 per site per 24h)
  • ·All ten score modules
  • ·PDF download with our branding
  • ·Multilingual reports

Starter

$29/mo

For freelancers running monthly reports.

  • ·Up to 5 tracked domains
  • ·50 audits per month
  • ·Fully white-labeled PDF reports
  • ·Custom brand kit (logo, color, footer)

Pro

$49/mo

For agencies, freelancers, and resellers.

  • ·Everything in Starter
  • ·Scheduled re-audits (weekly, biweekly or monthly)
  • ·Unlimited tracked domains
  • ·Email reports directly to clients

See plan limits and prices →

Which AI agents can use this MCP server

Any MCP-compatible client can connect over Streamable HTTP, or run the server locally as a stdio subprocess via the @metricspot/mcp-server npm package. Below are the agents we test against today.

  • •Claude Code: Anthropic's official CLI for Claude, with first-class MCP support.
  • •Cursor: the AI-first code editor. Add via ~/.cursor/mcp.json.
  • •Zed: the high-performance editor. Add via ~/.config/zed/settings.json.
  • •Claude Desktop: the macOS and Windows app, configured via claude_desktop_config.json.
  • •ChatGPT (Pro, Team, Enterprise): custom MCP connectors in Settings → Connectors.
  • •Gemini CLI: via the gemini CLI MCP configuration.

Running MetricSpot from an agent is one half of the AI-discoverability story. The other half is making your own site readable by ChatGPT, Perplexity, and Google AI Overviews: see our AI search visibility checks for the 11 rules MetricSpot runs on every page.

FAQ

What is MCP and why does an SEO tool need one?

Model Context Protocol is an open standard from Anthropic for connecting AI agents to external tools. Instead of asking your agent to scrape MetricSpot's UI or hand-write a REST client, you give it our MCP server URL and it discovers every tool, schema, and auth requirement automatically. For SEO work, that means an agent can audit a site, summarize the findings, fetch a PDF, and follow up on individual rules in a single conversation.

Do I need an API key to use the MCP server?

Only for 5 of the 6 tools. The run_audit_anonymous tool works with no key at all, rate-limited to 1 audit per IP per 24 hours: enough to try the server end-to-end. The other tools (run_audit, get_audit, list_audits, get_audit_pdf, get_organic_traffic) need a Bearer token. Sign up free at app.metricspot.com, then mint a key from Settings → API keys. The 10-key cap per account is plenty to spread across separate integrations: Claude Code on your laptop, a CI runner, a teammate's Cursor instance.

Which MCP clients are supported?

Any client that speaks Streamable HTTP MCP works. We test against Claude Code, Cursor, Zed, Claude Desktop, ChatGPT (Pro, Team, Enterprise), and the Gemini CLI. The protocol is open, so other clients should work too: if yours doesn't, please file an issue.

What transports does the server support?

Both transports are live today. Streamable HTTP at https://mcp.metricspot.com/mcp is the easiest path: one config line, no install. For agents that prefer to run the server as a local subprocess, install the @metricspot/mcp-server npm package and point your client at `npx -y @metricspot/mcp-server`.

What are the rate limits?

The anonymous tool is capped at 1 audit per IP per 24 hours. Authenticated tools count against your plan's monthly audit quota: the same number you see on the pricing page. There's no separate MCP quota or per-call surcharge.

What's the difference between anonymous and authenticated audits?

Same audit engine, same 91 rules, same 11 score modules. Authenticated audits add Core Web Vitals (via Google PageSpeed Insights), organic traffic data if you've linked Google Search Console, scheduled re-runs, and access to the branded PDF report. Anonymous audits are designed for quick checks and demos.

Can I run this in CI or a scheduled job?

Yes. The Streamable HTTP endpoint is stateless from the agent's perspective: you can call it from any CI runner, scheduler, or serverless function that can speak MCP. Use run_audit to queue, then poll get_audit until status is complete. Pair with get_audit_pdf if you want to email a branded report after each run.

Where do I file bugs or request new tools?

Open an issue at github.com/MetricSpot/mcp-server (preferred), or email info@metricspot.com. Tool requests from agents-in-production get prioritized. The MCP server is open source under MIT.

Stop writing SEO reports by hand.

Run an audit, brand the PDF, send to your client. In five minutes.

Start your first audit