---
title: "Agent skill · API docs · CheckAnySite"
description: "A SKILL.md file that teaches an agent when and how to use the API, for agents without MCP."
url: "https://docs.checkanysite.com/agent-skill"
---

For agents

# Agent skill

A SKILL.md file that teaches an agent when and how to use the API, for agents without MCP.

## Install it

The skills CLI installs the skill from our site into Claude Code, Cursor, Codex and other agents that support skills. The agent then needs your API key in the `CHECKANYSITE_API_KEY` environment variable:

```bash Terminal
npx skills add https://checkanysite.com
```

## What it teaches

The skill is called `checkanysite`. An agent reads its description to decide when to use it:

> Find out what a website is, with evidence for every answer. Tells whether it was built with an AI website builder (Lovable, Bolt, v0 and others), how ready it is for AI tools and what to fix, how popular it is, what it runs on, its trust signals, what it does and how old its domain is. Use it when the user asks about a specific website or domain. Needs a CheckAnySite API key in the CHECKANYSITE\_API\_KEY environment variable.

It teaches the agent to:

- Use the Lookup API first, and the Scan API only when there’s no stored result or it’s too old; and to use batches for several sites.
- Poll a scan until it’s complete, and present each finding with its evidence.
- Use the MCP server’s tools instead when the server is connected: check\_website, detect\_ai\_builder, get\_ai\_readiness, get\_popularity, get\_tech\_stack, get\_dns\_records, get\_trust\_signals, get\_domain\_history.
- Follow the rules: report signals, never verdicts; watch the credits; handle each error correctly.

## Where it is published

- The skill: [/.well-known/agent-skills/checkanysite/SKILL.md](https://checkanysite.com/.well-known/agent-skills/checkanysite/SKILL.md).
- The Agent Skills index, with the skill’s digest (Agent Skills discovery 0.2.0): [/.well-known/agent-skills/index.json](https://checkanysite.com/.well-known/agent-skills/index.json).
- Agents that support MCP don’t need the skill. [Connect the MCP server](https://docs.checkanysite.com/mcp) instead.

## Paste a prompt instead

For a single chat or an agent’s instructions, paste the following prompt. It gives the same rules as the skill in one message, and reads your key from the environment:

```text Prompt
Use the CheckAnySite API to find out what a website is, with evidence: traces of AI website builders, how ready it is for AI tools, its popularity in public ranked lists, the technology it runs on, its DNS records, trust signals with a score for how established it looks, what it is, and its domain history.

- Auth: send "Authorization: Bearer $CHECKANYSITE_API_KEY" on every call.
- Ask only for the checks you need: ai, readiness, popularity, tech, dns, trust, summary, history. In a query, ?checks= takes them separated by commas; in a body, "checks" takes an array. Leave it out to ask for all of them.
- Lookup API (our stored result, never a scan): GET https://api.checkanysite.com/v1/sites/{domain}?checks=ai,dns costs 1–2 credits a check, 3 for all. POST https://api.checkanysite.com/v1/sites/batch with {"domains": [...]} looks up as many as 50 sites.
- Scan API (when there is no result, 404 no_result, or you need a new one): POST https://api.checkanysite.com/v1/scans with {"domain": "...", "checks": ["ai"]} costs 1–6 credits a check, 15 for all. Then GET links.self every 3 seconds until "complete" is true; reading a scan is free.
- Send an Idempotency-Key header with every POST request, and the same one when you retry it.
- Errors look like {"error": {"code", "message", "docs"}}. GET https://api.checkanysite.com/v1/usage shows the credits left.
- Reference: https://docs.checkanysite.com/llms-full.txt · OpenAPI: https://api.checkanysite.com/v1/openapi.json

Report each finding with its evidence (the "evidence" lines) and link to the "resultUrl". Present findings as signals, not verdicts: never call a site safe, legitimate, a scam or fake.
```