SkillFoxx

Screaming Frog MCP: SEO crawl data for agents

Screaming Frog MCP

An unofficial headless MCP server for Screaming Frog: run crawls, export and analyze data with nine tools and no file write access

  • 96 ★ on the repository
  • Last commit
  • MIT
MCP server

Extended access

Why this level

  • `crawl_site` crawls a given site and creates real traffic to it
  • The `delete_crawl` tool permanently removes a saved crawl
All reasons and checks

bzsasson/screaming-frog-mcpAuthor:bzsasson

Install

In your terminal, with SkillFoxx CLI

npx skillfoxx add mcp/screaming-frog-mcp

Detects the agents on your machine, shows the access level and pins the version.

Other ways to install

Access and checks

How we review

Extended accessThe tool runs code or makes network calls and only reads external data.

  • `crawl_site` crawls a given site and creates real traffic to it
  • The `delete_crawl` tool permanently removes a saved crawl

This is third-party code. Review the repository files before installing.

What it does

The server drives the Screaming Frog SEO Spider command line and the saved crawl database, so an agent can run site audits and read their results without a GUI. It is a community project and is not affiliated with Screaming Frog. Since version 24 the app has an official MCP; this server is deliberately narrower: nine read and export tools, no script runner, no `npm install` and no filesystem write access. The tools: `sf_check` verifies the install and license, `crawl_site` starts a background crawl, `crawl_status` shows progress, `list_crawls` lists saved crawls, `export_crawl` exports CSV, `read_crawl_data` reads CSV with pagination and filters, `aggregate_crawl_data` computes counts and distributions ("how many pages return 404") server-side, `delete_crawl` removes a crawl and `storage_summary` shows disk use. Screaming Frog runs only when a tool needs it and does not stay up for the whole session. The allowed domains for `crawl_site` are set with `SF_ALLOWED_DOMAINS`.

Who it is for. SEO specialists and developers who run scheduled audits and CI checks on Screaming Frog and want to question an agent about saved crawls.

Good fit when

  • You need to know how many pages return 404 or a redirect without reading the whole export
  • A scheduled audit runs unattended and an agent reads the results
  • Old crawls take up disk and you need to see usage and delete extras
  • You need a CSV export of a crawl tab for spreadsheet analysis
  • A pre-launch audit skill uses crawl data

Not a fit when

  • You need visualizations, crawl comparison, screenshots and scripts: use the official MCP inside Screaming Frog 24
  • You have no paid license: the free version crawls up to 500 URLs and most features need a license
  • You want an interactive session with a crawl held in memory: this server keeps no state

Example request

List my saved crawls, take the latest one and count how many pages return 404 and which have the most internal links.

MCP

Transport
stdio
Authentication
not required

Environment variables

SF_CLI_PATH
Path to the Screaming Frog CLI, macOS path by default
SF_ALLOWED_DOMAINS
List of domains allowed for crawling
SF_CONFIG_DIR
Directory with `.seospiderconfig` files

Limitations

You need an installed Screaming Frog SEO Spider (tested on 23.x and 24.x, stated to work from 16), a valid license and the database storage mode. The internal database opens in one process only: close the Screaming Frog window before reading data or the server reports an error. On Linux and Windows set the CLI path with `SF_CLI_PATH`. The author advises installing with `uv tool install` and not running through `uvx`, because a cold cache can exceed the client's one-minute timeout. Exported CSV files are kept for one hour by default. `crawl_site` creates real traffic to a site, so set allowed domains.

How to disable. Remove the server block from the client configuration and uninstall the package with `uv tool uninstall screaming-frog-mcp`. Saved crawls remain in the Screaming Frog database.

FAQ

How is it different from the official Screaming Frog MCP?

The official server is built into the app since version 24 and offers about 29 tools, including visualizations and scripts. This server is small, with nine read and export tools, no file writes and no scripts, which suits unattended work.

Why does the server ask me to close Screaming Frog?

The internal crawl database opens in one process. While the app window runs, the server cannot read or export data and shows a clear error.

Do I need a Screaming Frog license?

Yes, for most features. The free version limits a crawl to 500 URLs, and the command-line mode, saving and export need a paid license.

How do I connect it to Claude Code?

Install the package with `uv tool install screaming-frog-mcp` and add the server with the absolute path to the executable, for example `~/.local/bin/screaming-frog-mcp`. Then ask the agent to call `sf_check`.

Can I limit which sites the agent crawls?

Yes. The `SF_ALLOWED_DOMAINS` variable takes a comma-separated list of domains, and `crawl_site` will accept only URLs from it.

Discussion

No questions or reviews yet.

MCP servers
Official

crawlseo MCP compares the latest 28 days of site metrics with the preceding 28 days, alongside results from the latest crawl.

MCP serverExtended access694

Google Search Console MCP

Google Search Console MCP Server for SEOs

An MCP server for Google Search Console: queries and pages with clicks and positions, URL indexing checks, sitemaps and ready SEO skills

MCP serverExtended access2.2K

All-In-One Free SEO Tool

All-In-One-Free-SEO-Tool

A self-hosted SEO suite with an MCP server and five skills for audits, keywords and site fixes in Claude Code and Cursor.

PluginBroad access126

The seo command and an MCP server for SEO work: page audits, Search Console, analytics and keywords. Sign-in through Google

CLIExtended access622

README in short

The README first explains the differences from the official MCP and unattended use, then lists requirements: Screaming Frog, a license, the database storage mode and Python 3.10. It describes the workflow (crawl, close the window, read data), two install routes, a table of nine tools, environment variables and the filter modes in `read_crawl_data`.

README badge

Report a problem