Brand extraction
Give it a domain. Get the logo, the real color palette, the fonts and the CSS design tokens back as JSON.
Learn MoreLive API, MCP server launching
AgentsRecon reads a website the way a developer would and hands back clean, structured data. Try it on any domain below. No sign-in needed for the preview.
6
API calls
5,000
Pages per crawl
2
SDK languages
100
Free pages a month
Features
Each one is a single request. Long ones run as async jobs with server-sent events or a webhook.
Give it a domain. Get the logo, the real color palette, the fonts and the CSS design tokens back as JSON.
Learn MoreOne URL in, clean Markdown out. Navigation, cookie banners and footers stripped, headings and links kept.
Learn MorePoint it at a site and get every page as Markdown. Runs as an async job with live progress or a webhook when done.
Learn MoreCapture a page at several viewport sizes in a single request, for visual QA, archives and design review.
Learn MoreClassify a business from its website into NAICS and SIC industry codes, no human labeling.
Learn MorePoint it at a PDF URL and get the text back, so reports, filings and manuals feed the same pipeline as web pages.
Learn MoreThe MCP server exposes the same calls as tools, so Claude Code, Claude Desktop or Cursor can look a site up on its own.
claude mcp add --transport http agentsrecon \ https://mcp.agentsrecon.com/mcp \ --header "Authorization: Bearer agr_YOUR_KEY"MCP setup for three clients
Pull a client's logo, colors, fonts and tokens straight from their live site before design starts.
See howTurn a list of domains into records with an industry code, a logo, brand colors and a short description.
See howCrawl a docs site or knowledge base to Markdown and feed it to your embedding or fine-tuning pipeline.
See howExport every page of an old site to Markdown, with a page count up front, before moving to a new CMS or stack.
See howFrom $19/mo for 5,000 pages. Or skip the account and buy one site as Markdown from $9.
Kick the tires. MCP included, rate-limited.
100 pages / month
Side projects and small agents.
5,000 pages / month
Production workloads.
50,000 pages / month
High volume, priority queue.
250,000 pages / month
I kept rebuilding the same scraper glue for client sites. AgentsRecon is that glue as an API, with an MCP server so an agent can use it without any of my code.
Read the storyCommon questions
One successful scrape of one URL is one page. A crawl counts each page it returns. Brand and styleguide calls use weighted credits because they fetch and analyze more than one asset.
No. Quotas reset each billing month. If you regularly use far less than your plan, the Free or Starter plan is the better fit, and you can change plans at any time.
Yes, on every plan. The Free plan is rate-limited. The MCP server launches alongside the public API release, and it exposes the same calls as tools, so a client like Claude Code can use them without extra glue code.
Yes. The tool on the home page runs a brand lookup on any public domain with no account. It currently returns clearly labelled sample data until the public preview endpoint ships.
Enter a domain on the home page, see how many pages we found, and pick the tier that covers it: up to 50 pages for $9, up to 250 for $29, up to 1,000 for $79. You pay once, sign in with a Titanium Labs magic link, and you get a ZIP of Markdown files, one per page.
No card. All endpoints. Add the MCP server when your agent needs it.