FAQ
Every question we hear,
in one place.
Grouped by audience. Answered as straight as we can manage. If something still isn't clear, the contact page is for that.
General
Is this AI?
Yes. We use modern vision-language models to read receipts the way a person would, return them as structured fields, and tag every value with a confidence score. The schema stays stable, so the underlying model can be upgraded as the field improves without breaking your integration.
Do I need to be a developer to use this?
No. The dashboard handles everything: drag a receipt in, get clean rows out, export to CSV or sync to your accounting tool. The API is there for teams who want to automate, not a requirement.
How accurate is it?
We don't publish a single accuracy number, because it depends on receipt quality, language, and merchant. Instead, every field comes back with a self-reported confidence score from 1 to 99. Route low-confidence rows for review, auto-approve high-confidence ones.
Is there a free trial?
No free tier and no trial. The floor is $5/month for 50 receipts. We charge enough to keep the service fast and reliable without rate-limiting you. Cancel anytime in the customer portal.
How is this different from a generic OCR API?
OCR returns text. We return data: structured items with variants, per-line tax, payment method with last four, multi-currency, language, confidence, and a one-paragraph summary. Versioned at /v1 so the schema stays stable. Per-receipt pricing with hard caps so you never get a surprise bill.
Can I delete my data?
Yes. Three retention modes per receipt: keep everything (default), delete the original file but keep the structured data, or delete both once your webhook confirms receipt. Set a workspace default, override per upload.
For developers
Sync or async?
Both. The API can return the structured data on the same response when extraction completes quickly, or you can fire-and-forget and listen for the signed webhook. Pick whichever fits your stack — the JSON shape is identical.
How do you handle retries?
Webhooks retry with exponential backoff on any non-2xx response. Each delivery includes a delivery id you can use to dedupe. Failed deliveries surface in the dashboard's delivery log with the response body for debugging.
Are there rate limits?
Yes, per workspace, scaled to your plan tier. The API returns 429 with a Retry-After header when you exceed them. Your dashboard shows current usage in real time. Hard plan caps return 402 instead.
How big can files be?
Single images up to typical phone-photo sizes. PDFs scale by page count. Multi-page documents queue and complete asynchronously via webhook. Hard limits and exact sizes live in the API docs because they evolve.
How do I test in development?
Live and test API keys are separate. Test keys use the same endpoints with sample receipts and don't count against your plan. Webhook signatures use a separate test secret so you can verify locally without rotating prod credentials.
What model is under the hood?
Modern vision-language models. We don't publish the specific provider because the model is a configurable component — what stays stable is the schema and the API surface. Upgrades roll out behind feature flags so accuracy improves without shape changes.
Do you have OpenAPI / SDKs?
Yes. The OpenAPI 3.1 spec is published at api.ripceipt.com/openapi.json. Our typed SDKs for JS/TS, Python, and Go are generated from it, so the method shapes match the REST endpoints exactly.
How is the schema versioned?
URI versioning at /v1. We add fields as backward-compatible additions only — no field renames, no type changes. Breaking changes get a new /v2 with at least twelve months of notice and an overlap period.
Does it work as a tool for AI agents?
Yes — it's designed for it. The single-POST endpoint maps cleanly to a function-calling tool: drop the OpenAPI schema into Claude Code, the Anthropic Agent SDK, MoltBot pipelines, OpenAI tool-calls, or LangChain. Confidence scores per field give the agent something to branch on. See /for/agents for the full pattern.
Pricing and billing
Why no free tier?
AI extraction has real per-receipt cost. We charge enough to keep the service fast and reliable without rate-limiting you. The floor is well below what most teams burn on a single hour of manual data entry.
What happens when I hit my cap?
The API returns 402 Payment Required, and the dashboard shows a one-click upgrade prompt. No silent overages, no metered surprises. Upgrade and the next call goes through; your existing data stays put.
Can I change tiers mid-cycle?
Yes. Upgrades take effect immediately, with prorated billing for the remainder of the cycle. Downgrades take effect at the end of the current period so you keep what you already paid for.
Do you offer annual discounts or enterprise pricing?
Yes to both. Every tier has an annual price that runs cheaper than paying month-to-month — flip the cycle toggle above to see your tier's annual rate. Enterprise covers custom volume pricing, SSO/SCIM, custom SLAs, and a security-review packet; book a call from the Enterprise section above.
How are receipts counted?
One receipt equals one successful extraction. Failed extractions and webhook retries do not count. Multi-page receipts count as one receipt regardless of page count.
Still have a question?
Drop us a line. We answer email like the small team we are.