Agent Checkout Preview

Early preview · The full platform is on its way

Agent Checkout

Payments intelligence for agentic commerce

An India-hosted LLM, customised for commerce. Grounded in the protocols, not guessing at them.

India-hosted LLM · Mumbai

Ask anything. Payments and commerce is where it is strongest.

No signup. Answers come from the documents below, running on our own hardware in Mumbai.

Build on it

Get a free API key, right here

No approval and no waiting. Your key appears on this page in seconds and works immediately against an OpenAI-compatible endpoint, so an existing OpenAI client needs two lines changed.

  • agentcheckout-1 answers payments and commerce questions from the documentation, with sources.
  • agentcheckout-1-chat is a general model for your own app: your system prompt, any subject.
  • 30 requests a minute. Runs in Mumbai. Nothing leaves India.

Full reference, examples and limits on the API docs.

What you can ask it

Shop and pay

Ask for a product and it searches a real catalogue, quotes the fetched price, and works the EMI out in the backend. Ask it to spend beyond its mandate and it declines, showing which check stopped it.

Find the best offer

Give it a product and a card and it returns the bank offer, cashback or no cost EMI worth most, with the saving and the effective price computed outside the model. Preview offer rates are illustrative and every answer says so.

Mandate intelligence

Score a UPI Autopay book for which debits are likely to fail, why, and when to present them again. Scoring runs in code; the history it runs over is sample data in the preview.

Integrate and onboard

API, webhook and error-code questions answered from the documentation, with sources. Onboarding questions answered in plain language rather than as a rejection code.

Powering merchant agents

Same model, same infrastructure. Different catalogs, different mandate scopes.

Customised for commerce

Grounded in around 1,200 documents across P3P, Grantex, ACP, UCP, AP2, x402 and a reference set on the Indian payment rails, with commerce tools and payment policy built around it. Duplicates and boilerplate are stripped before indexing. Hybrid retrieval, regression tested. Answers cite their source, and say when a source is unverified.

India-hosted

The LLM runs in Mumbai on our own hardware. No external API call. No data leaving the country.

Safe by architecture

Mandate scope and spend limits are enforced in code, outside the model. The model proposes, the backend decides. It will decline a purchase that exceeds its limit.