topic hub

Keeping AI agents and bots under control

Two AI-security problems arrived at once. Inside your product, agents now take real actions — calling APIs, deleting records, spending money — and a single bad tool call can do real damage. Outside it, a wave of AI crawlers scrapes your site, and the worst of them ignore robots.txt entirely.

agent·shield puts a firewall in front of agent actions, forwarding safe traffic instantly and holding destructive calls for a human to approve. robot·guard handles the other side, identifying and blocking the AI bots that don't play by the rules. Together they cover both directions of the same trust problem.

guides

frequently asked

Why can't I just trust an AI agent's tool calls?

An agent's actions are only as predictable as the model behind them, and a prompt injection or a confident mistake can turn a routine task into a destructive one. A guardrail layer lets safe actions through instantly while pausing irreversible ones — delete, pay, send — for a human to confirm.

Do AI crawlers actually respect robots.txt?

The major identifiable crawlers (GPTBot, ClaudeBot, CCBot, Google-Extended, PerplexityBot) publish their user-agents and state they honour robots.txt, but it's a voluntary request, not an enforced wall. A firewall is what stops the bots that ignore it or hide their identity.