Conv2pdf

What conv2pdf MCP server does

The conv2pdf MCP server gives an AI agent file-conversion and PDF-editing tools through the Model Context Protocol. It accepts local file paths, sends files to the conv2pdf API when processing is required, and writes returned results to the computer. Supported source formats include Word, Excel, PowerPoint, OpenDocument, Apple Pages, RTF, TXT, CSV, PNG, JPG, WebP, GIF, TIFF, HEIC, and EPUB. Numbers spreadsheets and Keynote presentations are not supported by the documented conversion tool.

The PDF toolset covers several common document workflows. Agents can convert text-based PDFs to editable DOCX files, render pages as PNG or JPG files in a ZIP archive, combine PDFs in a specified order, retain selected page ranges, rotate every page, add page numbers or a short text watermark, reduce image resolution, set or remove passwords, and manage printing or copying restrictions. A separate tool converts HEIC or HEIF photos to JPG.

How it works

Each operation takes a file path and normally creates a new file beside the source. Existing files are not overwritten; generated names receive an operation suffix or a numeric disambiguator. Callers can provide an output path, but the server refuses to proceed if that destination already exists. After the result reaches the local machine, the API deletes the remote job. If cleanup fails, the service deletes files after one hour.

PDF-reading tools can receive a known password for protected input files, and their outputs do not retain that password. Merging is the exception: protected PDFs must be unlocked before they are merged. Page ranges can use forms such as 1-5,7,10-12, and watermarks are limited to 50 characters.

The API allows 20 conversions per minute per key. File limits are 10 MB on the Dev plan and 200 MB on paid plans. Conversion attempts that are rejected after the API reads the file still count against quota; invalid arrivals, invalid settings, service failures, and get_quota do not.

Setup and configuration

The conv2pdf MCP server requires Node.js 22 or newer and the CONV2PDF_API_KEY environment variable. Create an account at conv2pdf.com, create an API key in the Developer dashboard, and keep the key available because it is shown only once. The key begins with cpdf_live_.

Run the published package over stdio with npx -y conv2pdf-mcp, passing the environment variable through the MCP client configuration. Claude Code can add it with claude mcp add; Claude Desktop and Cursor can use a JSON MCP server entry. Claude Code also has an optional repository plugin that stores the credential in its system credential store and provides a skill describing tool order and cost.

Without a key, the process starts and exposes its tool list, but tool calls explain that CONV2PDF_API_KEY must be configured. The server does not require a local conversion dependency; Node.js runs the package and communicates with the API.

Tools and capabilities

The conv2pdf MCP server exposes these operations:

• Convert supported office files, images, text files, or EPUB files to PDF.
• Convert HEIC or HEIF photos to JPG and text-based PDFs to DOCX.
• Export PDF pages as PNG or JPG files inside a ZIP archive.
• Merge, extract, rotate, number, watermark, compress, protect, and unlock PDFs.
• Query the API plan, usage, and limits with get_quota.

Limitations and notes

Processing is not entirely local: files used for conversion are sent to conv2pdf, although the README states that processing runs on OVHcloud servers in France and does not transfer data outside the EU. The service checks file content rather than relying only on extensions. Scanned or unreadable PDFs may be refused, and a PDF-to-Word conversion requires real text. Cancelled calls stop the transfer and leave no output file. Rate-limit retries occur at most twice for delays of 30 seconds or less; otherwise the returned delay must be observed.