The MCP server is live in production

Your AI agent audits your site on its own

Connect auditai.cc as a tool in Claude, Codex or Cursor. Type “audit mysite.com” and the agent hands you a link to the full report, graded A+ to F — without leaving the conversation.

How to connectSee the tools
No API keyNo sign-upReport in Polish and English

What it looks like

One sentence in chat, a finished report back

The agent decides when to call the tool. You just write normally, like a human.

1

You ask for an audit

A site address in a sentence is enough. The agent picks up the intent and calls audit_url.

2

The agent waits for you

The audit is asynchronous — it takes about 2–3 minutes. The agent polls get_audit_status and tracks the progress.

3

You get a report link

A full visual report: grade, categories, concrete fixes. The link is public, so you can pass it on.

Claude · connector: auditai.cc

Audit kwiaciarnia-zielona.pl for me, AI readiness wise.

Starting the audit.

audit_url({ url: "https://kwiaciarnia-zielona.pl", locale: "pl" })

get_audit_status → ANALYZING · step 2 of 5

Done. The report is waiting behind this link.

auditai.cc/raport/c3f9a1

Tools

Two tools, zero configuration

The server exposes exactly what it takes to order an audit and pick up the result.

audit_url

Orders an audit of a public page. Returns the identifiers the agent uses to track progress.

Input

url required
Address of the page to audit.

locale optional
pl or en. Defaults to pl — the report is written in that language.

Output

auditId — identifier used to poll the status

publicId — identifier used in the report address

cached — whether the result comes from a recent audit

reportUrl — only when cached — a ready link straight away

statusHint — what the agent should do next

get_audit_status

Checks the progress and, once finished, hands back the report address.

Input

auditId required
The same identifier audit_url returned — not publicId.

Output

status — audit stage: initial PENDING, intermediate stages (CRAWLING, ANALYZING, SCORING, … — the set may grow) and the terminal states COMPLETED and FAILED

progress — progress in percent

currentStep — the step running right now

estimatedRemainingMs — estimated time left

siteType — detected site type

reportUrl — after COMPLETED: /raport/<id>, in English /en/raport/<id>

The audit is asynchronous. audit_url returns immediately, while the result shows up after roughly 2–3 minutes. If the site is unreachable, you learn it from the status, not from a failed call.

Connecting

You paste one address and that is it

The server needs no login and no key. Configuration is a single field in your client's settings.

Claude

web and desktop app

  1. Open Customize → Connectors
  2. Click “+”, then Add custom connector
  3. Paste the address https://auditai.cc/api/mcp
  4. Confirm with Add — the tools show up in the conversation

This path applies to the Pro and Max plans. On Team and Enterprise an organization connector is added by the Owner under Organization settings → Connectors, and a team member then only clicks “Connect”.

Codex

configured in ~/.codex/config.toml

  1. Open the file ~/.codex/config.toml
  2. Add the server entry — no authentication is needed
[mcp_servers.auditai] url = "https://auditai.cc/api/mcp"

Cursor

configured in .cursor/mcp.json

  1. Create .cursor/mcp.json in the project or ~/.cursor/mcp.json globally
  2. Add the server under the mcpServers key
{ "mcpServers": { "auditai": { "url": "https://auditai.cc/api/mcp" } } }
Quick test for developers
curl -s -X POST https://auditai.cc/api/mcp \
  -H 'Content-Type: application/json' \
  -H 'Accept: application/json, text/event-stream' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Discoverability

Agents find us on their own

The server describes itself in standard files that AI clients fetch automatically.

Server card

application/mcp-server-card+json

A SEP-2127 compliant server card: schema, the name cc.auditai/audit and the endpoint address. The tool list is fetched separately, from the live tools/list.

auditai.cc/api/mcp/server-card

AI catalog

Catalog of AI services available on the domain.

/.well-known/ai-catalog.json

agents.json

Server entry for agents browsing the domain.

/agents.json

llms.txt

Description of the site and its tools for language models.

/llms.txt

tools/list

Full input schemas straight from the server.

POST /api/mcp

Before you ask

Four things people ask most often

Do I need an account?

No. The server requires no login and no API key. You open the report with the link the agent returns.

Can I audit someone else's site?

Yes, as long as it is publicly available. The audit reads the page the same way any search engine does. A malformed address is rejected immediately, at call time.

What language is the report in?

Polish by default, English when locale: "en" is set. The agent matches the language of your conversation.

How is this different from the audit on the site?

Not at all in substance — same engine, same report. The difference is the route: you order it from a chat instead of a form.

Connect it and type “audit my site”

Two minutes to configure, then an audit is one sentence away in chat. Prefer not to connect anything? Run it here, on the site.

Connection guideAudit on the site
2–3 minutesNo API keyFull A+ – F report