Serpel

What the Serpel MCP server does

The Serpel MCP server exposes SEO information from projects available in a Serpel account. It is intended for asking an AI assistant about website health, rankings, search traffic, technical crawl findings, competitors, and visibility in AI-generated answers. Results are read-only: the tools do not modify a website.

Project-level data includes audit scores, tracked keyword positions and changes, Search Console or Bing metrics, crawl issues, SEO recommendations, landing-page conversion performance, and the relationship between codebase routes, metadata, and live pages. The server can also report citations and mentions in ChatGPT answers and Google AI Overviews.

How it works

The remote endpoint is https://serpel.app/mcp, using Streamable HTTP. Clients that support OAuth can open a browser for first-time authentication. Another option is a Serpel API token sent in an Authorization: Bearer <token> header. The token needs mcp:read and the read permissions for the data the client should access.

Tools only return information from projects the authenticated account can access. List responses contain no more than 25 rows. Background operations can be checked with get_job_status, which reports a job's state and result.

The Serpel MCP server can return interactive cards for keyword rankings, crawl issues, keyword opportunities, and AI visibility in applications that support MCP Apps, including ChatGPT and Claude. Other clients receive the same information as text. Cards do not load external resources and only open links to the Serpel dashboard.

Setup and configuration

For Claude Code, add the remote server with:

claude mcp add --transport http serpel https://serpel.app/mcp

Cursor can use this configuration in .cursor/mcp.json:

{
  "mcpServers": {
    "serpel": { "url": "https://serpel.app/mcp" }
  }
}

VS Code with GitHub Copilot accepts the server in .vscode/mcp.json or through its MCP gallery. ChatGPT and Codex can connect through their connector settings or by installing the plugin. Gemini CLI can install the plugin directly from the Serpel GitHub repository. A local checkout can be loaded into Claude Code with claude --plugin-dir ./serpel-plugin.

The repository contains manifests and skills for supported clients; it does not contain server implementation code or hooks. Disconnecting a client is handled from Serpel's dashboard under Settings, API, and connected apps.

Tools and capabilities

The Serpel MCP server provides these tools:

• list_projects returns accessible project IDs and domains.
• get_project_overview reports audit scores, rankings, AI visibility, tasks, and next steps.
• get_keyword_rankings shows tracked terms, positions, and changes.
• get_search_performance returns clicks, impressions, CTR, and average position from Search Console or Bing.
• find_keyword_opportunities finds queries ranking from positions 4 through 20 with impressions.
• analyze_competitors compares visibility and stored competitor results.
• get_crawl_issues groups the latest crawl findings by severity.
• get_ai_visibility reports citations and mentions in ChatGPT and Google AI Overviews.
• get_seo_recommendations provides prioritized actions with impact, effort, and evidence.
• get_codebase_overview maps routes and metadata to live pages from a codebase snapshot.
• get_conversion_performance reports sessions, conversions, and conversion rate by landing page.
• research_keywords returns keyword ideas and metrics.

Limitations and notes

research_keywords uses credits and is not offered in ChatGPT. In other clients, it requires an active Starter or Pro plan, the workspace setting “Agents can spend credits,” and available daily and monthly allowances. The server only reads accessible project data, and every list is capped at 25 rows. Client-specific presentation differs: supported MCP Apps may show cards, while clients such as Claude Code, Cursor, Codex, and Gemini CLI receive text.