My AI agent got me banned from Resy. Another agent got me reinstated.

Brian Distelburger, the 47-year-old cofounder of AI performance review startup Windmill, got his Resy account deactivated after a personal AI agent scraped the reservation site for a table at 4 Charles. He turned around and tasked a second agent with drafting a reinstatement email. It worked within two days.

Distelburger runs a multi-agent system housed on a Mac Mini using the Hermes framework. One agent, named Concierge, was given his Resy credentials and told to check for openings at 4 Charles, a high-end New York steakhouse. He forgot to set guardrails. The agent polled the site constantly. Resy flagged the bot traffic and deactivated his account within 48 hours. It never made a single reservation.

“You don’t realize how important it is to have a Resy account in New York City until you don’t have one,” he said.

Distelburger emailed Resy four or five times apologizing and explaining what happened. He got no response, or a one-sentence reply. Someone told him accounts are reinstated after a month. He waited. Nothing.

Then he gave his primary agent, Hannah – named after his executive assistant of 15 years – a new task: draft an email that would maximize the chance of reinstatement. Hannah wrote a professional multi-paragraph letter that walked through exactly what he had done, why it was a mistake, and that he would not repeat it. The email referenced Resy’s terms of service, his long history as a Resy customer, and his American Express Business Platinum membership.

He sent it. Resy responded within two days and reinstated the account.

“I was so excited, relieved, and certainly learned my lesson: guardrails for agents,” Distelburger said. “They can really go wild.”

He now checks with Windmill’s engineering team before letting agents interact with external services. Most companies have not yet defined a strategy for agent traffic, he said. Until they do, users who build their own bots risk losing access to the very platforms they need.