Install Screaming Frog SEO Spider on a server or CI machine; on Linux, use the provided .deb or .rpm package or run via the headless JAR
Run a crawl from the command line using the --headless flag with arguments: --crawl <start-url> for a standard crawl, --output-folder to specify the export directory, and --save-crawl to persist the crawl session
Add --export-tabs to specify which data tabs to export (e.g., internal:all, response_codes:all, page_titles:all) and --export-format to set the output format (csv is default)
Use --config to pass a saved configuration file that pre-sets crawl settings such as JavaScript rendering mode, custom extraction rules, crawl speed, and authentication headers
Parse the CSV exports programmatically for downstream analysis; common automation targets include duplicate titles, missing meta descriptions, broken links (4xx), and redirect chains
Known gotchas
A valid Screaming Frog license is required for crawls exceeding 500 URLs; the free version silently stops collecting data at the 500 URL limit without an error, producing incomplete exports that appear complete
JavaScript rendering mode (using the internal Chromium renderer) increases memory and CPU usage substantially and reduces crawl speed; disable it unless the site requires JavaScript for critical content
The --output-folder path must exist before the crawl runs; Screaming Frog does not create the directory automatically and will fail silently or with an obscure error if the path is missing
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp