PDF to Structured JSONTables + Sections + Lists in one API call
Upload any PDF or pass a URL. Get clean nested JSON that includes heading hierarchy, paragraphs, list blocks, and machine-usable table data. Pricing starts at $0.05/page.
Why devs switch from enterprise parsers
Existing document-AI vendors optimize for procurement cycles, not indie shipping speed.
1. Overpriced usage tiers with minimum spend and annual lock-ins.
2. Output is often inconsistent across scanned pages and mixed layouts.
3. Setup friction blocks quick experiments in side projects and internal tools.
This product keeps pricing predictable and output schema stable so parsing can be a utility, not a project.
What You Get
Purpose-built extraction for developer workflows: less cleanup code, fewer brittle regex patches, faster time to production.
Hierarchical JSON
Outputs nested sections with heading levels, paragraphs, list groups, and metadata you can map directly to your schema.
Usable Table Objects
Returns each table as `headers` + `rows` arrays so you can push directly into analytics, ETL, or vectorization jobs.
Scanned PDF Recovery
When native text extraction is weak, pages are rendered and analyzed with Claude vision to preserve structure from image-only docs.
Built for Pipelines
Simple API surface, URL ingestion support, deterministic response format, and predictable pay-per-page pricing.
Simple Pricing
Pay as you go for sporadic workloads, or lock in a monthly bundle if you process docs continuously.
Pay As You Go
Best for variable usage and early-stage products
- No subscription required
- Great for low-volume and bursty workloads
- Direct upload or URL ingestion
- Structured JSON with table + hierarchy extraction
Builder Subscription
For teams running document pipelines daily
- Includes 1000 pages / month
- Automatic access token refresh via checkout
- Ideal for recurring ingestion jobs
- 4-6x cheaper than typical enterprise alternatives
FAQ
How accurate is table extraction?
Native-text PDFs preserve cell boundaries and are parsed directly. Scanned tables route through Claude vision, then normalized into row/column JSON.
Can I process large batches?
Yes. The API is optimized for pipeline usage: POST PDFs or URLs, then consume a deterministic JSON structure for downstream automation.
What counts as a billable page?
Every page processed counts once, whether parsed from embedded text or scanned with vision fallback. Rate is fixed at $0.05/page.
Do I need to create an account?
No account required. Checkout grants a cookie-based access token immediately after payment so you can start processing right away.