LibreCrawl MCP

Self-hosted SEO crawler for Claude — Screaming Frog-level audits, zero per-crawl cost.

Wraps LibreCrawl as a Claude MCP server. Give Claude the ability to fully audit any website — broken links, canonical issues, image alt text, orphan pages, Core Web Vitals, Schema.org, GSC errors — all running on your own server.

vs Screaming Frog

What LibreCrawl MCP gives you

LibreCrawl is a powerful self-hosted SEO crawler. On its own you get a web UI and raw crawl data. This MCP layer turns it into a fully automated Claude-native audit engine — one sentence to Claude gets you a complete site audit, saved as a Markdown report, ready to act on.

19 tools, zero context-switching

Claude gains 19 specialised SEO tools. Ask in plain English — Claude picks the right tool, runs it, interprets the output, and gives you a fix plan. No dashboards to navigate, no exports to download, no manual cross-referencing.

• librecrawl_audit() — one call: crawls the site, runs site-level checks, generates a structured Markdown report with a prioritised fix checklist. Done.
• Persistent authenticated session across all tool calls — no re-login between steps
• Every crawl saved to SQLite with full history via librecrawl_list_crawls()

Report engine — 35+ checks, 17 sections

LibreCrawl's raw JSON export gets transformed into a structured audit report covering everything Screaming Frog covers, plus things it doesn't:

Beyond crawl data — 6 specialist tools

GSC integration

Connect any Google Search Console MCP (we recommend mcp-gsc) and Claude will pull your real indexing errors — 404s in sitemap, server errors, soft 404s, crawl blocks — and append them to the audit report with fix hints. One conversation, complete picture.

Quick Start — Your First Audit

This guide takes you from zero to a full SEO audit report in under 10 minutes.

Step 1 — Install (1 command)

curl -fsSL https://raw.githubusercontent.com/adityaarsharma/librecrawl-mcp/main/install.sh | bash

What happens:

• LibreCrawl Docker image is pulled and built (~5–8 min, first run only)
• Python MCP server is installed and registered with PM2
• Both services start automatically and survive reboots

You'll be asked for a Google PageSpeed Insights API key (optional but recommended — free at console.cloud.google.com, 25k req/day).

Verify everything is running:

pm2 status librecrawl-mcp          # MCP server — should show "online"
docker ps | grep librecrawl        # LibreCrawl — should show "healthy"

Step 2 — Add to Claude

Claude Desktop — edit ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "librecrawl": {
      "type": "http",
      "url": "http://127.0.0.1:5081/mcp"
    }
  }
}

Claude Code — edit ~/.claude/settings.json:

{
  "mcpServers": {
    "librecrawl": {
      "type": "http",
      "url": "http://127.0.0.1:5081/mcp"
    }
  }
}

Restart Claude Desktop after saving. In Claude Code, the MCP is live immediately.

Step 3 — Run your first audit

Open Claude and type:

Audit https://example.com and give me a full SEO report

Claude will:

1. Crawl the entire site (runs in background, progress visible in pm2 logs librecrawl-mcp)
2. Run site-level checks (robots.txt, sitemap, HTTPS, www redirect)
3. Analyse all 35+ SEO checks across every crawled page
4. Save a Markdown report to ~/librecrawl-reports/example.com-YYYYMMDD-HHMMSS.md
5. Return a summary of the top issues with a prioritised fix checklist

For large sites (1,000+ pages), the crawl can run up to 2 hours. To cap it:

Audit https://bigsite.com — limit to 500 pages

Step 4 — Read your report

The report is saved at ~/librecrawl-reports/. Each report has:

Step 5 — Add GSC errors (optional)

Install a Google Search Console MCP (mcp-gsc is recommended), then ask:

Audit https://example.com and include GSC indexing errors

Claude will merge your real GSC coverage errors — 404s in sitemap, soft 404s, server errors, crawl blocks — into the audit report automatically.

Other things to ask

Check Core Web Vitals on the top 10 pages of example.com

Does example.com have schema markup? What rich results is it missing?

Run an internal links analysis — which pages are orphans?

Quick site health check for example.com — no crawl needed

Remote server setup (optional)

For local installs see the Quick Start above. To expose your LibreCrawl MCP over Nginx (e.g. on a VPS so your whole team can use it):

Claude config (via mcp-remote):

{
  "mcpServers": {
    "librecrawl": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://your-domain.com/librecrawl/mcp"]
    }
  }
}

Nginx location block for remote:

location /librecrawl/ {
    proxy_pass              http://127.0.0.1:5081/;
    proxy_http_version      1.1;
    proxy_set_header        Host $host;
    proxy_read_timeout      600s;
    proxy_buffering         off;
    proxy_cache             off;
    chunked_transfer_encoding on;
}

Add Google Search Console (optional but recommended)

LibreCrawl MCP can merge GSC coverage errors into your audit report via librecrawl_append_gsc_section. To use it, connect a GSC MCP server to Claude first.

Recommended: mcp-gsc (AminForou)

The best community GSC MCP — 876+ stars, maintained, supports OAuth and service account auth.

Install:

pip install mcp-search-console
# or with uvx (no pip needed):
# uvx mcp-search-console

Option A — OAuth (interactive, easiest):

1. Create a project at console.cloud.google.com
2. Enable Google Search Console API
3. Create OAuth 2.0 Client ID → Desktop app → download credentials.json
4. Add to Claude config:

{
  "mcpServers": {
    "gsc": {
      "command": "uvx",
      "args": ["mcp-search-console"],
      "env": {
        "GOOGLE_CREDENTIALS_FILE": "/path/to/credentials.json"
      }
    }
  }
}

First run opens a browser for Google auth. Token is cached.

Option B — Service Account (automation-friendly):

1. Create a service account at console.cloud.google.com → IAM → Service Accounts
2. Download the JSON key file
3. Add the service account email as a user in your GSC property (Settings → Users and permissions)
4. Add to Claude config:

{
  "mcpServers": {
    "gsc": {
      "command": "uvx",
      "args": ["mcp-search-console"],
      "env": {
        "GOOGLE_APPLICATION_CREDENTIALS": "/path/to/service-account.json"
      }
    }
  }
}

Using GSC with audit reports

Once connected, Claude can pull GSC errors and merge them into any audit:

"Audit uichemy.com, include GSC indexing errors"

Claude will:

1. Run librecrawl_audit("https://uichemy.com") → gets report_path
2. Pull GSC coverage errors via the GSC MCP
3. Call librecrawl_append_gsc_section(report_path, gsc_data) → adds a GSC section to the report

The GSC section includes: indexing errors with fix hints, crawl errors, manual actions, and a prioritised fix checklist.

Tools (19 total)

Architecture

Claude
  │  MCP (streamable-http)
  ▼
Python MCP server (port 5081)
  │  REST API
  ▼
LibreCrawl Flask app (port 5080, Docker)
  │  Headless crawl (Playwright + Chromium)
  ▼
Target website

Stack: LibreCrawl · FastMCP · httpx · uvicorn · PM2 · Docker

Manage services

# Status
pm2 status librecrawl-mcp
docker ps | grep librecrawl

# Logs
pm2 logs librecrawl-mcp
docker logs librecrawl --tail 50

# Restart
pm2 restart librecrawl-mcp
docker restart librecrawl

# Stop
pm2 stop librecrawl-mcp
docker stop librecrawl

Configuration

Related

• LibreCrawl — the crawler this wraps
• mcp-gsc — Google Search Console MCP (876 stars, recommended)
• Model Context Protocol

License

MIT — use freely, attribution appreciated.

Built by Aditya Sharma