# Knocklog: who's knocking?

AI agents read your website every day: crawlers training models, AI search engines, and assistants fetching pages for someone who asked about you. Some of them are not who they say they are. Knocklog shows which agents visit, verifies each one, and turns that traffic into data you can use.

## Check your site for free

Open https://knocklog.ai/?site=yourcompany.com, or call `GET https://api.knocklog.ai/v1/scan?domain=yourcompany.com`, or use the `scan_site` MCP tool. The check reads robots.txt, the homepage, llms.txt and the markdown version the way an agent would, and returns a 0–100 score with what to fix.

## Measure real agent visits

Create a free project (`POST https://api.knocklog.ai/v1/projects` with `{"domain": "yourcompany.com"}` or the `create_project` MCP tool). You get a send key, a report key and two ways to connect:

- **Cloudflare Worker**: real time, sees signatures, passes visitors to your site unchanged.
- **Access logs**: any host, including cPanel shared hosting, and brings in history.

The report shows agents by class and name, which were verified and which were spoofed, the pages agents read most, and how many people arrived from AI assistants.

## How we tell a real agent from an impostor

1. **Cryptographic signature**: Web Bot Auth signatures checked against the operator's published keys.
2. **Published IP ranges**: OpenAI, Anthropic, Perplexity, Google, Apple and Microsoft.
3. **Reverse DNS**: for search crawlers that document it.

Verified means proven to be the agent it claims. Spoofed means proven not to be. Everything else is unverified.

## For AI agents

MCP endpoint: `https://api.knocklog.ai/mcp`. In Claude Code: `claude mcp add --transport http knocklog https://api.knocklog.ai/mcp`. Summary for agents: https://knocklog.ai/llms.txt

## Privacy

We never store full IP addresses, query strings or cookies from your visitors. Our scanner identifies itself as KnocklogScanner/0.1 and reads only public pages.