OctoBoost SEO MCP Server

Expose the OctoBoost SEO API as Model Context Protocol (MCP) tools so agents can audit websites with compact, structured results instead of fetching and parsing raw HTML.

Get your free API key at octo-boost.com. New accounts include free credits to try every tool.

What This Server Does

octoboost-mcp-server gives an MCP client three core capabilities:

• discover available analyzers with list_analyzers
• crawl a domain for relevant URLs with scan_domain
• run full audits with analyze

It is built for agent workflows that need SEO and AI-visibility signals inside the reasoning loop without spending thousands of tokens on raw page content.

Who It's For

This server is a good fit for:

• developers building MCP-enabled products, assistants, or internal automation
• teams using MCP clients such as Cursor or Claude Desktop and wanting SEO tooling via config only
• AI agent workflows that need token-efficient site audits, progress updates, and structured outputs they can reason over

It is less useful if you want a general SEO learning guide or a raw HTML scraping tool. The main value here is compact audit output for automated workflows.

Why Use This Instead Of Raw Scraping?

Running SEO checks directly in an LLM context is expensive. OctoBoost moves the heavy lifting to the API and returns only the signals an agent needs to decide what to do next.

• Token-efficient: structured results instead of raw HTML
• LLM-friendly: scores, flags, and diagnostics instead of prose parsing
• Credit-aware: responses include credits used and credits remaining
• Predictable errors: 401 for invalid or expired keys, 402 for exhausted credits

Quick Start

1. Get an API key from octo-boost.com.
2. Add the server to your MCP client config.
3. Call list_analyzers to verify the connection.

{
  "mcpServers": {
    "octoboost-seo": {
      "command": "npx",
      "args": ["-y", "octoboost-mcp-server"],
      "env": {
        "OCTOBOOST_API_KEY": "your-api-key"
      }
    }
  }
}

Common config locations:

• Claude Desktop: ~/Library/Application Support/Claude/claude_desktop_config.json
• Cursor: Cursor MCP settings
• OpenClaw: ~/.openclaw/mcp.json

Core Workflow

Most agent flows follow this pattern:

1. list_analyzers
   -> learn categories and available checks

2. scan_domain { domain: "acme.com" }
   -> collect relevant URLs for the audit

3. analyze { urls: [...] }
   -> run a full audit with SEO score + GEO/AEO score for each URL

Tools Overview

list_analyzers

Returns available analyzer keys, categories, and weights. Call this first so an agent knows what it can run.

• no input required
• current categories include seo, accessibility, ux, performance, and geo
• weights returned reflect your personal setup (see Analysis Setup below)

scan_domain

Crawls a domain and returns SEO-relevant URLs.

| Parameter | Type | Default | Description | | --------------------- | -------- | -------------- | ----------------------------------------------------------------------- | | domain | string | — | Domain or URL to scan | | maxPages | number | server default | Maximum pages to crawl (capped server-side at the batch analysis limit) | | excludePatterns | string[] | [] | URL patterns to skip | | respectRobotsTxt | boolean | true | Honor robots.txt

... View full README on GitHub