---
name: agentsiteaudit
description: Audit any public web page and get one structured JSON report with 69 checks, evidence and exact fixes, then verify the fixes later. Paid per call with x402, 0.01 USDC on Base.
---

# AgentSiteAudit

Use this when you need to know whether a website is healthy, secure, findable, accessible, fast and ready for AI agents, or to confirm that fixes worked.

## Audit a page

`GET https://agentsiteaudit.com/v1/audit?url=<page>`

- An unpaid request returns HTTP 402 with a PAYMENT-REQUIRED header: x402 version 2, exact scheme, 10000 units of USDC on Base (eip155:8453). Pay, then retry the same request with a PAYMENT-SIGNATURE header.
- The answer is a JSON report. Read `summary.score`, then `issues` (most important first), then each check's `evidence` and `fix`.

## Verify fixes

`GET https://agentsiteaudit.com/v1/audit?url=<page>&previous=<audit_id>`

- Returns a fresh report plus `comparison`: fixed, improved, still_failing, worse, new, not_rechecked and new_checks, with the score before and after.
- Same price. An unknown audit_id, or one for a different site, returns 400 before any payment.

## Rules

- NOT_TESTED is never a pass. Read `not_tested_reason`.
- Check IDs are permanent. Ignore fields and IDs you do not recognise.
- You are charged only when an audit actually runs.
- The agents category (Agent readiness) is shown but not yet counted in the score.

## MCP

Connect any MCP client to `https://agentsiteaudit.com/mcp/` (Streamable HTTP). Tools: `audit_site`, `verify_fixes`, `list_checks`. 5 free audits a day per connection; after that, use the paid API above.

## More

- Check registry: https://agentsiteaudit.com/v1/checks.json
- OpenAPI: https://agentsiteaudit.com/v1/openapi.json
- Guide: https://agentsiteaudit.com/llms.txt
- Paid resources: https://agentsiteaudit.com/.well-known/x402
- Verified listing with a settled payment: https://payapi.market/api/agentsiteaudit

## Customer reaction check

- `GET https://agentsiteaudit.com/v1/reaction?url=https://example.com/` costs $0.05 via x402 (same flow as /v1/audit).
- Eight AI-simulated customer perspectives read the homepage. You get up to 8 findings, each with a category, severity, the personas affected, evidence (a checked word-for-word quote, or what is missing), why it matters and a suggested fix. No score.
- Not charged when the page cannot be read, has too little text, the AI step fails, or fewer than 2 findings can be backed by evidence.
- To check fixes, add `&previous=<reaction_id>`. If the page text has not changed you get the earlier findings back free. If it has, `comparison` lists resolved, improved, unchanged, worsened, new and variation. Treat `variation` as noise, not a change.
- Simulated perspectives, not real customers. Homepage text only.