# Knocklog

> Knocklog shows which AI agents visit a website, verifies whether each one is who it claims to be, and reports that traffic for marketing, governance and sales. It is free to start and usable entirely by agents over MCP.

Knocklog classifies every request as a training crawler, AI search crawler, assistant acting for a person, agentic browser, coding agent, search engine or other bot, and verifies identity with Web Bot Auth signatures, the operators' published IP ranges and reverse DNS. "Spoofed" means proven false; anything that cannot be checked is "unverified".

## For agents

- [MCP server](https://api.knocklog.ai/mcp): Streamable HTTP. Tools: scan_site, classify_request, list_known_agents, create_project, get_install_plan, import_logs, get_agent_traffic. Project tools take the project key as `api_key` or an `Authorization: Bearer` header.
- [API index](https://api.knocklog.ai/): REST endpoints (`/v1/scan`, `/v1/classify`, `/v1/agents`, `/v1/projects`, `/v1/events`, `/v1/logs`, `/v1/traffic`).
- [Known agents](https://api.knocklog.ai/v1/agents): the directory of AI agents with operator, class, robots.txt token and verification method.

## Typical flow for a user's site

1. `scan_site` with the user's domain: readiness score and what to fix (robots.txt rules for AI agents, JavaScript-only content, llms.txt, markdown).
2. `create_project` with the domain: returns a send key (`kl_ing_…`), a report key (`kl_adm_…`, keep private) and install steps. Show both keys to the user once.
3. Connect traffic: a Cloudflare Worker (real time) or access-log import with `import_logs` (any host, includes history).
4. `get_agent_traffic` with the report key: agents by class and name, verified vs spoofed, pages agents read, visits referred by AI assistants.

## Pages

- [Home and free site check](https://knocklog.ai/): also served as markdown with `Accept: text/markdown`.