For AI coding

Run SEOCode inside Cursor and Claude.

Q: Does my code leave my machine?

No. The MCP server runs locally over stdio as a child process of your editor. Your files are analysed on your machine and nothing is sent to a server.

Q: Is it free?

Yes, completely. The MCP server is free and local. It runs the full ruleset. The paid Studio plan is for the GitHub App enforcing rules on every pull request across your team.

Q: Which editors work?

Any MCP client: Cursor, Claude Desktop, Claude Code, and OpenAI Codex. Most use the same JSON config block; Codex uses a TOML entry in ~/.codex/config.toml. ChatGPT's own connectors need a hosted remote server, so the local server does not plug into the ChatGPT app.

Q: Does it honor .seocode.json?

Yes. It walks up from the file you audit to find your .seocode.json and applies the same rule disables and severity overrides the CLI and GitHub App use.

AI assistants write code fast, and quietly ship broken metadata, invalid JSON-LD, and blocked AI crawlers. The SEOCode MCP server lets you audit the file you're editing right in your AI chat - locally, free, with the exact fix - before it ever reaches a pull request.

Set it up → Install the GitHub App

Free and local. Works in Cursor, Claude Desktop, Claude Code, OpenAI Codex, and any MCP client. Nothing leaves your machine.

# ask your AI assistant, in any file:
"audit this file for SEO issues"

How it works

Your AI assistant, plus a real SEO engine.

MCP (Model Context Protocol) lets your editor call external tools. The SEOCode server runs as a local process and exposes the same deterministic engine the CLI and GitHub App use - so the checks never hallucinate.

Add one config block

Point your MCP client - Cursor, Claude Desktop, Claude Code, or OpenAI Codex - at the seocode-mcp server. It launches locally over stdio - no account, no network, no dashboard.

Ask in chat

Type "audit this file for SEO issues" while editing. Your assistant calls the tool, which parses the file and runs the full SEO and GEO ruleset against it.

Get findings and fixes

Issues come back grouped by severity, each with the exact fix. Ask for a 1-click fix and it returns a deterministic replacement you can apply on the spot.

The tools

Two tools, one engine.

The server exposes exactly two tools. Both are thin wrappers over the engine you already trust - no new logic, no guessing.

`audit_file`

Runs the full ruleset against one file. Framework-aware (Next.js, Astro, SvelteKit, Remix, Vue, plain HTML), so it won't false-flag dynamic metadata. Returns issues grouped Critical / Warning / Info with the fix for each.

`suggest_fix`

Returns a deterministic 1-click code replacement for a specific issue. Only rules that can safely rewrite the source produce a replacement; others return written guidance - it never invents a fix.

Local preview, not the safety net

The MCP audits the file you ask about, when you ask. The GitHub App enforces every rule on every pull request across your team, automatically. Use the MCP while you code; use the App so nothing slips through review.

Setup

One config block.

Works in any MCP client - Cursor, Claude Desktop, Claude Code, and OpenAI Codex. Most use the same JSON block (Claude Desktop's claude_desktop_config.json, Cursor's MCP settings, or a project .mcp.json):

{
  "mcpServers": {
    "seocode": {
      "command": "npx",
      "args": ["-y", "-p", "@qobi/seocode", "seocode-mcp"]
    }
  }
}

OpenAI Codex uses TOML instead - add to ~/.codex/config.toml:

[mcp_servers.seocode]
command = "npx"
args = ["-y", "-p", "@qobi/seocode", "seocode-mcp"]

Reload your editor after adding it. Prefer a local build? Point command at node and the built dist/src/mcp/index.js. Full steps in the docs.

FAQ

Questions, answered.

Does my code leave my machine?

No. The server runs locally over stdio as a child process of your editor. Your files are analysed on your machine; nothing is sent anywhere.

Is it free?

Yes, completely - free and local, running the full ruleset. The paid Studio plan is for the GitHub App enforcing rules on every pull request across your team.

How is this different from the GitHub App?

The MCP audits the one file you are editing, when you ask. The GitHub App reviews every pull request for the whole team automatically and can block a merge until issues are fixed. The MCP is the local preview; the App is the team safety net.

Which editors work?

Any MCP client - Cursor, Claude Desktop, Claude Code, and OpenAI Codex. Most use the same JSON config block; Codex uses a TOML entry in ~/.codex/config.toml. ChatGPT's own connectors need a hosted remote server, so the local server does not plug into the ChatGPT app.

Does it honor my .seocode.json?

Yes. It walks up from the file you audit to find your .seocode.json and applies the same rule disables and severity overrides the CLI and GitHub App use.

Get started

Catch SEO bugs while you type.

Add the MCP server to your editor for the local preview, then install the GitHub App so the same rules run on every team pull request.

Set it up → Install the GitHub App