BrowsePDF
Get a free key

MCP server · Document API

PDF tools for your AI agent

BrowsePDF is a hosted MCP server and REST API that lets Claude, Cursor, or any agent work with PDFs — convert, merge, split, redact, OCR, and compose multi-step pipelines. Point your agent at one URL, no infrastructure to run.

Get a free API keyRead the API docs

Endpoint: https://browsepdf.co/api/mcp · Bearer auth · Streamable HTTP

Connect in 60 seconds

  1. Grab a free API key — email, no card.
  2. Add the server to your MCP client's config, with your key as a Bearer token.
  3. Ask your agent to do something with a PDF. That's it.

Cursor · VS Code (native remote MCP)

{
  "mcpServers": {
    "browsepdf": {
      "url": "https://browsepdf.co/api/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY" }
    }
  }
}

Claude Desktop (mcp-remote bridge)

{
  "mcpServers": {
    "browsepdf": {
      "command": "npx",
      "args": [
        "-y", "mcp-remote", "https://browsepdf.co/api/mcp",
        "--header", "Authorization: Bearer YOUR_API_KEY"
      ]
    }
  }
}

17 tools your agent can call

Upload a file to get a handle, run operations that return new handles, then download — or chain it all in one run_pipeline call.

  • upload_from_urlPull a file from a URL into a document handle
  • convertOffice / HTML → PDF
  • mergeCombine PDFs into one
  • splitSplit by page ranges
  • organize_pagesReorder, delete, duplicate, rotate pages
  • watermarkStamp a text watermark (keeps text)
  • page_numbersAdd running page numbers
  • fill_formFill (and optionally flatten) AcroForm fields
  • extract_textGet the text layer, with word/block boxes
  • renderRasterize pages to PNG/JPEG
  • redactPermanently burn out regions
  • compressShrink file size (lossy)
  • ocrScanned PDF → searchable PDF + text
  • get_metadataPages, sizes, encrypted/signed/fillable (free)
  • get_download_urlShort-lived download URL for a result
  • run_pipelineCompose several ops as a DAG in one call
  • get_jobPoll an async job (convert / ocr / pipeline)

Composable by design

Agents work best when one call does the whole job. Handles flow between steps with no intermediate downloads, and slow steps (convert, OCR) run async with a job to poll.

// One call: convert a Word file, watermark it, then compress.
run_pipeline({
  inputs: { src: "<document id from upload_from_url>" },
  nodes: {
    conv: { op: "convert",   in: "src" },
    mark: { op: "watermark", in: "conv", options: { text: "CONFIDENTIAL", tile: true } },
    small:{ op: "compress",  in: "mark", options: { level: "balanced" } }
  },
  output: "small"
})

Built for agents, not just humans

  • URL-in, URL-out

    Pass files by URL and get results as download URLs — no base64 juggling in the context window.

  • Deterministic

    Same input + operation + options → byte-identical output, so agent runs are reproducible.

  • Self-describing

    Clear tool names, an OpenAPI 3.1 spec, and an llms.txt — agents can discover and call it zero-shot.

  • Free to start

    A self-serve key with free credits and no card. Metered, rate-limited, and honest about cost.

OpenAPI specllms.txtllms-full.txtFull API reference

Questions

What is the BrowsePDF MCP server?

It's a hosted Model Context Protocol (MCP) server that gives an AI agent a full set of PDF tools — convert, merge, split, watermark, redact, compress, OCR, extract text, render, fill forms, and multi-step pipelines. It's a thin adapter over the BrowsePDF Document API, so auth, metering, and limits are enforced in one place. The endpoint is https://browsepdf.co/api/mcp.

How do I add PDF tools to Claude, Cursor, or Cline?

Get a free API key, then add the server to your client's MCP config. Clients with native remote MCP (Cursor, VS Code) take a url + Authorization header; stdio-only clients like Claude Desktop connect through the mcp-remote bridge. Both snippets are on this page.

Is there a free tier?

Yes. A self-serve key comes with free credits and no credit card. Metadata reads are free; light operations cost 1 credit; page-heavy ops (render, OCR, compress) scale per page. You only pay if you exceed the free credits.

Do I need to host or install anything?

No. The server is hosted at https://browsepdf.co/api/mcp — you just point your agent at the URL with your key. Files can be passed by URL (upload_from_url) and results come back as short-lived download URLs, so it's URL-in, URL-out.

Which PDF operations are supported?

convert, merge, split, organize_pages, watermark, page_numbers, fill_form, extract_text, render, redact, compress, ocr, plus get_metadata, upload_from_url, get_download_url, run_pipeline, and get_job — 17 tools. Operations return new document handles you can chain, or compose into a DAG pipeline in a single call.

Can it also be used as a plain REST API?

Yes. Every tool is a REST endpoint under /api/v1, with an OpenAPI 3.1 spec and an llms.txt for agents. Use whichever fits — the MCP server and the API share the same key, metering, and limits.

Give your agent PDF superpowers

Free key, hosted server, 17 tools. Point your agent at a URL and go.

Get a free API key