MCP server · zero-install or open-source

pseolint MCP server

Audit programmatic SEO sites for SpamBrain and answer-engine risk without leaving your AI assistant. Paste one URL for the hosted remote server, or run the open-source stdio package locally. Both speak the Model Context Protocol, so they drop into Claude, Cursor, and any MCP-capable client.

Recommended · zero install

Remote server

One hosted endpoint — no Node, no npx, no local setup. Serves the three read-only audit tools. Anonymous use is rate-limited; a free API key raises your limits.

https://pseolint.dev/mcp

Cursor & clients with native remote support

Add to your MCP config (e.g. ~/.cursor/mcp.json):

{
  "mcpServers": {
    "pseolint": {
      "url": "https://pseolint.dev/mcp"
    }
  }
}

Claude Desktop (stdio-only) — via the mcp-remote bridge

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "pseolint": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://pseolint.dev/mcp"]
    }
  }
}

Using an API key (higher limits)

Create a key under dashboard → account and send it as a Bearer token. Native config:

{
  "mcpServers": {
    "pseolint": {
      "url": "https://pseolint.dev/mcp",
      "headers": { "Authorization": "Bearer YOUR_KEY" }
    }
  }
}

Bridge form: npx -y mcp-remote https://pseolint.dev/mcp --header "Authorization: Bearer YOUR_KEY"

Open source · power users

Local stdio server

Runs on your machine from the open-source @pseolint/mcp package. Adds the AI-orchestrated pseolint_orchestrate_audit tool, which produces a fix manifest of concrete, paste-able patches.

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

The orchestrate tool calls a paid LLM provider — set ANTHROPIC_API_KEY(or another supported provider key) in the server's environment. The other three tools run without it.

Remote vs. local

  • Remote: zero install, 3 read-only tools, anonymous (rate-limited) or keyed.
  • Local: all 4 tools incl. orchestrate, fully offline, unlimited.

The tools

  • pseolint_audit_site — verdict + risk + findings
  • pseolint_explain_score — why, quick wins first
  • pseolint_check_page_technical — per-page checks
  • pseolint_orchestrate_audit — fix manifest (local)

Frequently asked

What's the difference between the remote and the local (stdio) server?
Same engine, different delivery. The remote server is hosted at https://pseolint.dev/mcp — you paste one URL into your MCP client, nothing to install, and it serves the three read-only audit tools. The local stdio server runs on your machine via npx @pseolint/mcp; it also exposes the AI-orchestrated audit tool that produces a fix manifest with concrete patches. Use remote for the fastest start, stdio for the orchestrate tool or fully offline/local runs.
Is it free? Do I need an account?
The remote server's three read-only tools are free and need no signup — anonymous use is rate-limited per IP. Creating a free API key in your dashboard raises those limits. The open-source stdio package is free and unlimited. The orchestrate tool (stdio/CLI) calls a paid LLM provider with your own API key.
Which clients does this work with?
Any MCP client. Clients that support remote Streamable HTTP (e.g. Cursor) take the URL directly. Stdio-only clients like Claude Desktop reach the remote server through the mcp-remote bridge, or you can install the stdio package locally. Setup snippets for each are above.
What does the audit actually check?
pseolint runs 48 rules mapped to documented Google SpamBrain signals (thin content, doorway patterns, near-duplicate clusters, boilerplate ratio, scaled-content density, site-reputation abuse) plus answer-engine-optimization checks for AI Overviews, Perplexity, and ChatGPT search. It audits by template, returning a per-template verdict and a 0–100 risk score, not a flat per-URL list.
Is it safe to point it at any URL?
Yes. Every tool runs in SaaS-safe mode: a DNS-validated guard blocks requests to private/internal IP ranges (SSRF protection), fetch sizes are capped, and robots.txt is honored. The remote endpoint can't be turned into a proxy to internal hosts.
What are the four tools?
pseolint_audit_site (crawl + verdict + risk score + findings), pseolint_explain_score (why the verdict is what it is, quick wins first), pseolint_check_page_technical (per-page canonical/OG/JSON-LD/meta checks), and pseolint_orchestrate_audit (LLM-driven deep audit that emits a paste-able fix manifest — stdio/CLI only for now).

New to pseolint? Start with the free web tools or read the calibration methodology.

The MCP server exposes four tools over the Model Context Protocol: an async site audit that returns a per-template risk score and full findings JSON, a score-explanation tool that surfaces quick wins ranked by severity, a per-page technical check covering canonical tags, Open Graph, and JSON-LD, and an AI-orchestrated audit (stdio only) that produces a paste-able fix manifest. All tools run the same 48-rule engine as the web UI — no separate install or account required for the remote endpoint.

Sources