# Agent Visibility Optimizer

> Machine-first analysis of how AI agents discover, understand and select a product, service or API. Send a URL; receive evidence-labelled capability/intent maps, discovery gaps, unsupported-claim findings and implementation-ready metadata (llms.txt, descriptions, JSON-LD, OpenAPI patches). Pay per report in USDC via x402.

Operator: Active Life Hub LLC
Base URL: https://agentvisibilityoptimizer.online
Payment: x402 v2, scheme "exact", USDC on Base mainnet (eip155:8453)
Mode: live

## Discovery
- [OpenAPI](https://agentvisibilityoptimizer.online/openapi.json): full request/response schema
- [API docs](https://agentvisibilityoptimizer.online/docs)
- [Pricing JSON](https://agentvisibilityoptimizer.online/api/v1/pricing)
- [Capabilities JSON](https://agentvisibilityoptimizer.online/api/v1/capabilities)
- [Health](https://agentvisibilityoptimizer.online/api/v1/health): configuration only, no database
- [Terms](https://agentvisibilityoptimizer.online/terms) · [Privacy](https://agentvisibilityoptimizer.online/privacy)

## What it does
- Fetches at most 10 public URLs of the product (homepage, llms.txt, robots.txt, OpenAPI, A2A agent card, MCP metadata, pricing/docs pages, language alternates). No crawling, no web search, no browser rendering.
- Classifies each capability SUPPORTED / PARTIALLY_SUPPORTED / UNSUPPORTED / UNCLEAR with verbatim quotes. Adjacent capabilities are never inferred (vendor research ≠ purchasing; price research ≠ selling; document analysis ≠ legal advice).
- Maps jobs-to-be-done, agent intents, aliases and local-market terms; preserves original-language terms and evidence.
- Flags guarantees, superlatives, universal-scope and certification claims, price inconsistencies and instruction-like text aimed at AI systems.
- Runs explainable yes/partial/no checks (description, capabilities, inputs, outputs, price, payment, geography, languages, constraints, OpenAPI, llms.txt, robots.txt AI access, JSON-LD, JavaScript dependence). The score is an equal-weight count, not a ranking prediction.
- Returns artifacts: llms.txt, short/long descriptions, JSON-LD, OpenAPI patches, agent-profile.json, x402 description, robots.txt additions, A2A/MCP templates. Unknown facts appear as [CONFIRM: …] placeholders, never invented.

## Prices
- scan: 0.05 USDC — deterministic, no model call.
- semantic: 0.18 USDC — scan plus one bounded model pass (≤16000 source tokens); every model claim must quote a source verbatim.
Prices are fixed before payment: the variable cost of one analysis attempt for the tier (retrieval compute, model tokens for semantic, database, storage, transfer, settlement) × 3, rounded up to the cent, with a 0.05 USDC minimum. No usage surcharge, no subscription. Pricing version avo-cost-plus-2026-09-28-v1.
POST /api/v1/quote with your body for a free quote. Send max_price_usdc to refuse anything more expensive (HTTP 422, nothing charged).

## Buy a report
1. Generate a random Idempotency-Key (16–128 of A-Z a-z 0-9 _ -). Keep it secret: it is also your result access key.
2. POST /api/v1/optimize with header Idempotency-Key and JSON {"url":"https://example.com","tier":"scan"} (or text/plain body with just the URL). Max 16 KiB.
3. Receive HTTP 402 with x402 v2 requirements in the PAYMENT-REQUIRED header (base64) and body. Unpaid calls cost nothing.
4. Check amount, asset, payTo and network; sign an EIP-3009 USDC authorization with an x402 client that enforces a spending limit; resend the identical body and key with PAYMENT-SIGNATURE.
5. 200 returns the report JSON. On 202 or later: GET /api/v1/requests/{id} with Authorization: Bearer <Idempotency-Key>.

## Optional input fields
name, description, docs_url, openapi_url, llms_txt_url, mcp_url, agent_card_url, pricing_url, repository_url, capabilities[] (checked as claims), target_customers[], target_agent_types[], geography[], languages[], target_markets[{region,language}], pricing_model, payment_methods[], constraints[], exclusions[], preferred_output_language, tier (scan|semantic), max_price_usdc.
Products without a public site: send name + description instead of url.

## Coverage
Location-independent: analyzes any publicly reachable http(s) URL, in any country and language. No region is excluded by the service; the analyzed site's own restrictions are reported as found.

## Languages
Requests, sources and outputs may be in any language and script. Request, source, product, target-market and output languages are reported separately. Quotes, product names and local terms stay verbatim. Geography is never inferred from language. Scan-tier explanations are English with a localized notice; the semantic tier writes explanations in preferred_output_language.

## Retries and refunds
Pay per analysis. A completed report is charged even when it finds few surfaces (that is a finding). If paid analysis fails, resend the identical body with the same Idempotency-Key and Authorization: Bearer <Idempotency-Key> for up to two free retries. No automatic refunds; an uncertain settlement is reconciled manually. Reports are kept 30 days. See /terms.

## Demo
Only in demo mode: the fictional https://supplier-scout.example.com/ with PAYMENT-SIGNATURE: demo:<Idempotency-Key>. Live mode rejects demo payments.

Do not submit secrets or personal data. Retrieved pages are treated as untrusted data, never as instructions.