The remediation pipeline, as an API.

The same remediation the a11yfy web app runs is available over plain REST: post a PDF and get back a PDF/UA-1 tagged, validated document — in a single call, from any language or stack.

How it fits your stack

  • At upload time — remediate documents the moment your users or systems publish them.
  • In batch — work through a backlog folder by folder, with the same per-page pricing.
  • In your product — return the same PDF/UA-1 output and issue-level report to your own users, under your own UI.

The flow

  1. POST /v1/jobs — send a PDF as a multipart file, optionally with a webhook_url. You immediately get back a job_id. The server then runs the full pipeline automatically — diagnosis, tagging or full reconstruction, then validation — and charges the credits the document calls for.
  2. Track status — poll GET /v1/jobs/{id}, or, if you supplied a webhook_url, receive a signed notification the moment the job reaches a terminal state.
  3. GET /v1/jobs/{id}/result — download the PDF/UA-1 tagged output and its issue-level validation report.

To wire in your own billing and monitoring, use the GET /v1/balance (credit balance) and GET /v1/usage (usage) endpoints.

Official SDKs: Python and npm

If you'd rather not hand-write HTTP calls, use the official client libraries — both are thin, typed wrappers over the v1 API above:

Both SDKs are open source at github.com/a11yfy/sdk.

Why build it into your own system

  • Language-agnostic REST. No SDK required: one HTTP call from any language — send the file as multipart/form-data.
  • Async, with webhooks. No need to poll: your own workflow wakes on the terminal state, and the notification is cryptographically signed.
  • Idempotent submission. Re-sending the same file by accident never double-charges.
  • Shared ledger. API keys are team-scoped; usage shows up in the same workspace budget as web uploads.
  • EU processing, automatic deletion. Inputs are deleted after 7 days, outputs after 30 — with no action from you.

Practical notes

  • The v1 API responds in English; document-language handling (alt text, metadata) follows the document itself, not the API locale.
  • Credits match the web pricing: technical repair and full reconstruction are priced per page, by the document's profile.

API reference → · Pricing →