AI Agent Log Checker
Personal AI agents such as Meta Muse, Grok Bot and Hark now open websites for their users. Paste lines from your access log to see which agents visited, from how many IPs, and whether the operator documents the name. It runs in your browser and nothing is uploaded.
Summary
| Agent | Type | Evidence | Client IP | Requests |
|---|
The agents this checker recognises
Official means the operator's own page prints the token. Reported means Cloudflare's verified-bot registry or another public source lists it but the operator does not, so treat the name as likely rather than confirmed. Verified on 11 October 2026.
| Agent | Operator | Token | Type | Evidence |
|---|---|---|---|---|
| Grok Bot (GrokAgent) | xAI | grokagent | Personal agent | Reported |
| Meta Muse | Meta | muse- | Personal agent | Reported |
| Meta-ExternalTest | Meta | meta-externaltest | Integration test | Reported |
| Meta-ExternalFetcher | Meta | meta-externalfetcher | User-triggered fetch | Official |
| Hark Agent | Hark | harkbot | Personal agent | Official |
| Google-Agent | google-agent | Personal agent | Official | |
| ChatGPT-User | OpenAI | chatgpt-user | User-triggered fetch | Official |
| Claude-User | Anthropic | claude-user | User-triggered fetch | Official |
| Perplexity-User | Perplexity | perplexity-user | User-triggered fetch | Official |
| MistralAI-User | Mistral AI | mistralai-user | User-triggered fetch | Official |
| DuckAssistBot | DuckDuckGo | duckassistbot | User-triggered fetch | Official |
| ZipchatBot | Zipchat | zipchatbot | Shopping assistant | Official |
| Amzn-User | Amazon | amzn-user | User-triggered fetch | Official |
| GrokBot | xAI | grokbot | Crawler name | Reported |
| xAI-SearchBot | xAI | xai-searchbot | Crawler name | Reported |
Agents that leave no name in your logs
- OpenAI dots and ChatGPT Work's Cloud browser. Dots work from their own cloud computer and browser. OpenAI documents that the Cloud browser signs requests with Web Bot Auth and a
Signature-Agent: "https://chatgpt.com"header, and that Cloudflare, Akamai, HUMAN and Vercel list it as ChatGPT Agent. Read what OpenAI dots mean for your website. - Instinct and other texting agents. No public token or verified-bot entry as of 11 October 2026. They appear as ordinary browser traffic.
- Agents using the person's own computer. OpenAI says a dot can use the user's local browser when its cloud browser is blocked, and xAI lets Grok Bot route through a member's desktop. Those requests look like that person's normal browser.
For these, your CDN's bot analytics is the right place to look, not the access log. To see how an agent experiences your pages, run the AI agent readiness check.
Frequently asked questions
Are my logs uploaded anywhere?
No. The checker runs in your browser. Nothing you paste is sent to our servers or stored.
Which AI agents can it find?
Any of the 15 agent and user-fetcher tokens in the table on this page, including Grok Bot (GrokAgent), Meta Muse (muse-), Hark Agent (HarkBot), Google-Agent, ChatGPT-User, Claude-User and Perplexity-User. Each result links to its profile and shows whether the operator documents the token.
Why does it not find OpenAI dots?
Dots browse from their own cloud computer and do not send a token of their own. OpenAI documents Web Bot Auth signatures for ChatGPT Work's Cloud browser (Signature-Agent: "https://chatgpt.com"), and bot managers such as Cloudflare, Akamai and HUMAN list it as ChatGPT Agent. Standard access logs do not record those headers, so check your CDN's bot analytics instead.
Can a matching name be faked?
Yes. Anyone can copy a user-agent string. Several of these agents sign their requests with Web Bot Auth, defined in RFC 9421, so a CDN that verifies signatures can tell the real agent from a copy. The IP column helps you spot impostors: one real agent usually comes from a small set of addresses.
Should I block personal AI agents?
Usually not on public pages. A personal agent visits because one person asked it to do something on your site, such as compare, book or buy. Block or challenge it on paths where you do not support agent sessions, and do it at the firewall, because robots.txt is not documented as a control for most agents.
What log formats does it read?
Any line that contains the user agent, including the Nginx and Apache combined formats, where the user agent is the last quoted field and the client IP is the first field. If your CDN exports a Signature-Agent header, lines that include it are grouped as signed agents.