ai_env

Detects which AI coding agent (Claude Code, Codex, Cursor, Gemini, etc.) is running the current process, by inspecting agent-specific environment variables and filesystem markers.

Usage

To start, detect if an AI agent is running the process.

ai_env::is_ai_agent();

Or detect which agent is being used.

ai_env::detect_agent(); // Claude

And extract information about the agent environment.

use ai_env::get_environment;

if let Some(ai) = get_environment() {
	println!("Agent: {:?}", ai.agent);

	if let Some(id) = ai.id {
		println!("Self-reported: {}", id);
	}

	if ai.sandboxed {
		println!("Running in a sandbox");
	}
}

Additionally, detect whether the environment restricts network egress, e.g. the GitHub Copilot cloud agent firewall, Codex sandboxes with network access turned off, or local filtering proxies like the Claude Code sandbox.

use ai_env::{detect_agent, detect_network_policy, AiNetworkPolicy};

match detect_network_policy(detect_agent()) {
	AiNetworkPolicy::Disabled => println!("No network access"),
	AiNetworkPolicy::Filtered => println!("Egress restricted to an allowlist"),
	AiNetworkPolicy::Open => println!("Egress explicitly unrestricted"),
	AiNetworkPolicy::Unknown => println!("No signal"),
}

This is derived from environment variables only — no network I/O is performed. A Filtered network may still allow the hosts your program needs, so verify reachability of those hosts before acting on the signal.