AI agent directory
Who's reading your site? A field guide.
Every AI crawler and agent Punchline recognizes by name: who runs it, why it fetches, and how to check a request really is it. Generated from the registry Punchline's classifier uses, so it's exactly what we detect.
Fetches for a person asking now
Fetching for someone asking right now
- **ChatGPT-User** (OpenAI): verified by published IP ranges
- **Claude-User** (Anthropic): verified by published IP ranges
- **Perplexity-User** (Perplexity): verified by published IP ranges
- **MistralAI-User** (Mistral AI): verified by published IP ranges
Acts in a browser for a person
Acting for a person in a browser
- **ChatGPT agent** (OpenAI): verified by signed requests (Web Bot Auth)
- **Google-Agent** (Google): verified by published IP ranges
Builds a search or answer index
Building search and answer indexes
- **OAI-SearchBot** (OpenAI): verified by published IP ranges
- **Claude-SearchBot** (Anthropic): verified by published IP ranges
- **PerplexityBot** (Perplexity): verified by published IP ranges
- **Googlebot** (Google): verified by published IP ranges, reverse DNS
- **Bingbot** (Microsoft): verified by published IP ranges, reverse DNS
- **Applebot** (Apple): verified by published IP ranges, reverse DNS
- **DuckDuckBot** (DuckDuckGo): verified by published IP ranges
- **MistralAI-Index** (Mistral AI): verified by published IP ranges
Collects training data
Collecting training data
- **GPTBot** (OpenAI): verified by published IP ranges
- **ClaudeBot** (Anthropic): verified by published IP ranges
- **CCBot** (Common Crawl): verified by published IP ranges
Use it
Take the list with you
The same list as JSON, with user-agent patterns, IP range lists, reverse DNS suffixes and signature key directories: /agents.json. Registry version 2026-10-25.1.
Agents that aren't on this list are still detected, by behavior: HTTP clients, headless and automated browsers, cloud networks and machine-like request rates. See Concepts.