Skavio Processing MCP

Skavio exposes its Processing API to AI agents as a standard, paid Model Context Protocol server. A model or agent can upload a document or media file, get a price, run the work on Skavio's real processing pipeline, download the signed result, and settle it against prepaid Skavio credits.

OAuth 2.0 + PKCE for hosted clients (see OAuth and authentication)

This is not a mock, a wrapper or a second wallet. It calls the same Processing API and the same credit wallet that power the Skavio website, and it sells the same operations at the same verified cost-plus prices.

What you can do

TaskOperation idUnit
PDF to textpdf_to_textpage
Extract a page range from a PDFpdf_splitsource document
Merge PDFspdf_mergesource document
OCR a scan or imagedocument_ocrpage
Extract invoice / receipt fieldsinvoice_extractdocument
Transcribe audio or videoaudio_transcribestarted minute

Current prices are in pricing and as machine-readable data in pricing.json.

Start here

  1. Getting started — create a company, connect a client, run your first paid job.
  2. Authentication — create, rotate and revoke an MCP credential.
  3. OAuth 2.0 sign-in — interactive PKCE sign-in for Smithery and other hosted clients.
  4. Tools and workflow — every tool, argument and the exact paid sequence.
  5. Client configuration — copy-paste config for common MCP clients.
  6. Pricing — the price list and how credits work.
  7. Troubleshooting — error codes and how to recover.
  8. Privacy and security — data handling, isolation and limits.
  9. Integration examples — runnable Python and TypeScript clients.

Ground rules

enforced 25–40% gross margin band from verified, dated cost evidence. If that evidence is missing or stale, paid admission fails closed with a clear error.

never read, spend or download for another company.

forwards your MCP token upstream and never exposes the upstream key to you.

checkout. Credits are purchased only by a company administrator in the Skavio dashboard.