SmartPDF

What SmartPDF MCP server does

SmartPDF MCP server exposes PDF and photo-processing operations to AI agents through MCP. The repository describes 47 operations, including conversions between PDF, Word, PowerPoint, Excel, JPG, PNG, and image formats, as well as OCR, compression, merging, splitting, page editing, redaction, signatures, watermarks, passwords, and document metadata cleanup.

The available operations cover both document transformation and page-level changes. Agents can extract, delete, rotate, reorder, or crop pages; add page numbers or annotations; flatten forms and comments; and export PDF text in plain, block, or word-coordinate modes. Photo workflows can assemble JPG, PNG, WEBP, HEIC, TIFF, and BMP files into PDFs, optionally applying scanning, grayscale, resizing, and EXIF removal.

How it works

The hosted MCP endpoint is https://mcp.smartpdf.ru/mcp and uses Streamable HTTP. Authentication is supplied with an API key through either an Authorization: Bearer header or an X-API-Key header. Most MCP tools receive public URLs rather than local file paths and return a link to the processed file.

Operations expose parameters for common variations. For example, OCR accepts language codes and supports up to 100 pages per request; PDF-to-image tools accept 72, 150, or 300 DPI; and HTML-to-PDF supports A4, A3, or Letter in portrait or landscape orientation. The get_balance tool reports account balance, monthly spending, request pricing, and remaining included requests.

Setup and configuration

Create an account at SmartPDF, confirm the email address, and generate an API key beginning with sk_live_ in the API and MCP section. The repository states that email confirmation adds 300 ₽ to the API balance. The service is billed per request in rubles, with pricing and plan details provided through the developer portal.

Claude Code can install the repository plugin through its marketplace commands. Cursor has a repository plugin and an MCP configuration, or can connect directly to the hosted endpoint. Other documented configurations include Claude Desktop, VS Code, Codex CLI, Gemini CLI, Windsurf, Cline, n8n, and Make. For local files, the repository also includes an Agent Skill and a Python 3 script that calls the REST API; that workflow uses the SMARTPDF_API_KEY environment variable.

Tools and capabilities

The listed MCP tools support:

• Office, image, HTML, and PDF conversion in both directions where supported.
• OCR for Russian and English, including invisible text layers in scanned PDFs.
• Compression, merging, splitting, extraction, deletion, rotation, reordering, cropping, and page numbering.
• Text extraction, PDF editing, visual signatures, watermarks, annotations, and image placement.
• AES-256 password protection, password removal, flattening, and personal-data redaction.
• PDF-to-JPG and PDF-to-PNG exports as ZIP archives, plus Excel extraction from PDF tables.

Limitations and notes

MCP file inputs must be publicly accessible URLs, and the documented file limit is 100 MB. OCR is limited to 100 pages per request; larger documents should be split first. Cropping does not remove content outside the visible crop, so redaction should be used when data must be removed from the file. The service runs on servers in Moscow, uses ruble billing, and is governed by the service offer rather than the repository license. Repository files are licensed under MIT.