Machine-Readable Surfaces # Built for AI agents, not just browsers. Every way an agent -- a chat assistant, a crawler, a browser extension -- can read or act on this case file, in one place. Reports remain self-reported and unverified everywhere they're surfaced; see [/llms-full.txt](https://scammercasefile.com/llms-full.txt) for the exact framing agents should use. The surface area today: 6 read-only MCP tools behind 3 discovery endpoints, a Markdown twin for every case file plus 2 llms.txt documents (14 curated links), and 8 AI crawlers explicitly allowed in robots.txt — live since the site launched on 2026-07-20. ## Remote MCP server A real Model Context Protocol server -- JSON-RPC over the Streamable HTTP transport, not a mock. Exposes 6 read-only tools: search reports, look up one exact number, list by country or scam type, scam trends, and recent reports. Filing a report is intentionally not exposed as a tool -- it stays a human action through the /report form. ### [/mcp](https://scammercasefile.com/mcp) JSON-RPC endpoint POST JSON-RPC 2.0 requests here (initialize, tools/list, tools/call). Stateless -- no session required. ### [/mcp/server-card](https://scammercasefile.com/mcp/server-card) SEP-2127 Server Card Static pre-connection metadata (identity, transport, protocol versions) served as application/mcp-server-card+json, at the spec-recommended \/server-card location. ### [/.well-known/mcp/catalog.json](https://scammercasefile.com/.well-known/mcp/catalog.json) MCP Catalog The genuinely well-known discovery entrypoint -- lists every MCP server this domain hosts (one) and points to its Server Card above. ## In-browser tools (WebMCP) The same two read-only actions, registered with the page's own navigator.modelContext host via the WebMCP proposal -- for an agent already driving the browser, not a separate connection. ### navigator.modelContext.registerTool() Browser API (draft) Registered on every page load if a WebMCP host exists. No shipping browser exposes one by default yet -- this is forward-compatible, not yet reachable in production. ## Machine-readable page content Case file pages are also available as plain Markdown, for agents that would rather read prose than render a page. ### [/scammer/{number}.md](https://scammercasefile.com/scammer/1670982113.md) Markdown twin Same case file as /scammer/{number}, as a GFM Markdown document -- advertised via \ on the HTML page. Example linked. ### [/llms.txt](https://scammercasefile.com/llms.txt) Site summary Short, llms.txt-convention overview of what this site is and how it works. ### [/llms-full.txt](https://scammercasefile.com/llms-full.txt) Full reference The complete data model, URL structure, and explicit guidance for how agents should characterize report data (self-reported, unverified -- never "confirmed"). ## Crawler policy Standard discovery documents, with AI crawlers explicitly allowed rather than left to a default. ### [/robots.txt](https://scammercasefile.com/robots.txt) Crawler rules Explicit Allow rules for GPTBot, ChatGPT-User, PerplexityBot, ClaudeBot, anthropic-ai, Google-Extended, Bingbot, and CCBot. ### [/sitemap.xml](https://scammercasefile.com/sitemap.xml) URL index Every public page, regenerated from live D1 data on each request. ## The standards behind these surfaces Nothing above is invented here — each surface implements a published convention: the [Model Context Protocol](https://modelcontextprotocol.io/), the [llms.txt proposal](https://llmstxt.org/), the [WebMCP draft](https://github.com/webmachinelearning/webmcp), and the robots.txt rules standardized by [RFC 9309](https://www.rfc-editor.org/rfc/rfc9309.html). > "MCP (Model Context Protocol) is an open-source standard for connecting AI applications to external systems. … Think of MCP like a USB-C port for AI applications." — [modelcontextprotocol.io](https://modelcontextprotocol.io/) > "A proposal to standardise on using an /llms.txt file to provide information to help LLMs use a website at inference time." — [llmstxt.org](https://llmstxt.org/) ## Frequently asked questions ### Do I need an API key to use the MCP server? No. The MCP endpoint at /mcp is public, free, and unauthenticated -- connect and call tools directly. All tools are read-only. ### Can an agent file a report through MCP? No, by design. Filing a report creates a public record naming a phone number, so it stays a human action through the /report form. Every MCP tool is read-only. ### What's the difference between /llms.txt and /llms-full.txt? /llms.txt is a short, llms.txt-convention overview of what this site is and how it works. /llms-full.txt is the complete data model, URL structure, and explicit guidance for how agents should characterize report data. ### Which AI crawlers are allowed to index this site? /robots.txt explicitly allows GPTBot, ChatGPT-User, PerplexityBot, ClaudeBot, anthropic-ai, Google-Extended, Bingbot, and CCBot, rather than leaving AI crawling to a default. ## Been targeted by a scammer? File a report with the number, name, and what happened — it's published immediately and warns the next person who searches. [Report a scammer](https://scammercasefile.com/report)