Merge, split, rotate, compress — a PDF MCP server your AI agent calls directly
Paplume ships a standard MCP (Model Context Protocol) server. Claude, Codex, Cursor and any MCP-capable agent can merge, split, compress and rotate PDFs with one line of config — anonymous, free, no API key.
Endpoint:POST https://pdfsuite.zalize.com/mcp
{
"mcpServers": {
"pdf-suite": {
"type": "http",
"url": "https://pdfsuite.zalize.com/mcp"
}
}
}Standard protocol
Built on the official MCP TypeScript SDK: a stateless Streamable HTTP endpoint that plugs into any MCP client.
Zero friction
Anonymous calls — no signup, no API key. Each tool has a daily free per-IP quota drawn from the same IP pool as the website (with a stricter cap on the MCP side).
Failures refunded
Only successful operations consume quota; failed calls are refunded automatically (up to a daily refund allowance). Quotas reset at 00:00 UTC.
Honest limits
Rendering-based conversions (PDF to image / OCR) and AI features stay browser-only — where files are processed locally on-device.
Why let your agent call MCP directly?
Stay in the conversation
Merge, split and compress right inside Claude / Cursor — no switching to a website, no upload/download round-trips.
Composable
Multi-step PDF pipelines (split → watermark → compress) are chained by the agent automatically — one sentence is enough.
Zero setup cost
No API key, no signup — one JSON snippet connects you; refund semantics match the website (quota draws from the same IP pool, with a stricter MCP cap).
Setup (Claude / Codex / Cursor)
Add this entry to your MCP client configuration:
{
"mcpServers": {
"pdf-suite": {
"type": "http",
"url": "https://pdfsuite.zalize.com/mcp"
}
}
}Available tools (9)
| Tool | Use when you want to… |
|---|---|
| merge_pdfs | combine 2–10 PDFs into one |
| split_pdf | extract a page range (1-based, inclusive) |
| rotate_pdf | rotate all or specific pages by 90/180/270° |
| remove_pages | delete specific pages |
| compress_pdf | shrink file size (lossless structural compression) |
| add_watermark | stamp diagonal Latin text on every page |
| add_page_numbers | number pages bottom-center |
| edit_metadata | read / set / clear title, author, subject, keywords |
| flatten_pdf | bake AcroForm field values into page content |
Input conventions
- Each PDF input accepts base64 (file content) or url (https, fetched server-side).
- Limits: ≤10 MB per file, ≤200 pages. Non-PDF inputs are rejected.
- Successful calls return a text summary plus the result PDF as base64 — decode and save it.
Not offered over MCP (honest boundaries)
- Rendering-based conversions (PDF to images, grayscale, OCR) and AI translate/summarize/chat are browser-only, where files never leave your device.
- compress_pdf does not downsample images; for stronger compression use the website’s target-size compressor.
- edit_metadata clear mode resets standard info fields only; full XMP scrubbing is offered by the website’s metadata tool.
Prefer the website?
43 free tools that process files locally in your browser — no uploads.
Open the free tools