A2A, MCP and WebMCP: which interface should an agent use?
PDFMDTools exposes the same narrow public capability in several forms: finding relevant site guidance for a PDF or Markdown question.
| Interface | Best fit | What it does here |
|---|---|---|
| A2A | One agent calling another service | Accepts a text question and returns an immediate guide result. |
| MCP | An MCP-compatible desktop or server client | Provides the find_site_page tool. |
| WebMCP | A capable assistant inside the website | Exposes the same public lookup capability in the browser. |
Use A2A for agent-to-agent lookup
The A2A Agent Card describes two A2A 1.0 bindings. Use A2A when a calling agent needs a normal request-and-response lookup without depending on a browser page.
Use MCP for tool-oriented clients
The MCP Server Card points to /mcp. The available tool searches a bounded public page index. It does not open private files or make changes.
Use WebMCP inside a browser
WebMCP is useful when a compatible browser assistant is already on PDFMDTools. It can discover the page-level lookup tool without a separate integration.
What none of these interfaces do
They do not upload PDFs, run OCR, reconstruct table layouts, save a chat history, or export a document for a caller. For actual conversion, a person can use the browser PDF-to-Markdown tool or Markdown-to-PDF tool locally.