Anthropic via x402: turn supplier offers into a comparison
A Claude Sonnet supplier-comparison request through x402engine, including a live quote and an independently checkable commercial example.
- Written by
- Agentic Economy
- Published
We looked at how an agent could use an Anthropic model to organise supplier offers into a comparison. We checked x402engine's Claude Sonnet 4.6 route and obtained a free quote for a two-offer example. We've set out the request, price and independently checkable arithmetic so you can assess a first trial.
The useful work extends beyond adding up charges. A comparison should show which terms were supplied and which questions still need an answer. Our example is designed to check that distinction with a setup fee, two monthly rates and deliberately incomplete commercial terms.
The gateway lists POST https://x402engine.app/api/llm/claude-sonnet-4.6 for x402 purchases. Its service record identifies OpenRouter as the upstream provider. This differs from Anthropic's direct Messages API, so we use the gateway's request contract below. We obtained a quote without purchasing a model response.
A comparison you can check yourself
Consider two invented offers for the same service:
| Offer | Setup | Monthly | First 12 months |
|---|---|---|---|
| A | $2,000 | $300 | $5,600 |
| B | $0 | $500 | $6,000 |
These totals are calculated as setup plus twelve monthly payments. The example assumes the same currency and tax treatment and no additional charges. It establishes an answer you can verify before relying on a model to interpret messier offers.
The useful model contribution is organising the comparison and surfacing absent terms: included work, usage limits, contract period, cancellation, support and extra fees. It cannot infer those terms from a supplier's name.
Request a bounded result
Save the following as request.json:
{
"messages": [
{
"role": "user",
"content": "Compare these supplier offers. A: $2000 setup and $300 monthly. B: no setup and $500 monthly. Calculate the first-year total for each and list missing commercial terms."
}
],
"max_tokens": 400
}
Get a free quote before buying the response:
curl https://x402engine.app/api/llm/claude-sonnet-4.6/quote \
-H 'Content-Type: application/json' \
--data-binary @request.json
On 7 September 2026, this returned $0.057724 USD, with a 400-token output budget. The price is for this request and budget. The gateway API schema documents the paid and quote routes; the client instructions cover x402 payment.
We made the free quote call only. The comparison table above is our calculation, not output from a paid Claude call.
Define the result before accepting it
A successful response should reproduce the two totals, identify that A is $400 cheaper over the stated first year and separate missing information from supplied facts. A recommendation to sign with A would go beyond the evidence: the offers may differ in scope, service levels or exit terms.
For a real comparison, extract the relevant commercial passages from the offers and retain their document references alongside them. Ask the model to tie each comparison field to that supplied text. If a condition is absent, the result should say so. This lets the operator turn unknowns into specific questions for the suppliers.
Use ordinary calculation for the arithmetic when integrating the workflow. The model can help interpret and organise text; a deterministic calculation gives the totals a checkable basis. A trial containing one deliberately incomplete offer will show whether the output invents missing terms or leaves them visible.
The business boundary
This service can produce a comparison draft. It does not negotiate terms, verify a supplier's performance or commit the business to a purchase. The person responsible for procurement still evaluates the tradeoffs.
The workflow becomes more useful when the output is a consistent comparison record rather than another long narrative. Keep fields for price basis, included scope, exclusions and questions to resolve. The same format can support the next round of offers without prescribing the eventual decision.
Start with the two-offer example, check the arithmetic and missing-term behaviour, then introduce representative excerpts from the business's own procurement work.