Hey, founder here 👋 I built Relay because of something that happened to me directly. I was building a custom Python agent and gave it database access to handle a routine task. One bad instruction later, it exposed live credentials. No warning. No checkpoint. The agent just decided, then acted. I found out after the fact. What surprised me most while fixing it wasn't the complexity — it was how easy it was to break. Agents don't need to be hacked. They just need one ambiguous instruction and access to a tool that can do real damage. I built relay to solves this problem: every tool call should pass through a policy gate before it executes. Allowed or blocked. Logged either way. No agent code changes needed after the initial wrap. That's Relay. A runtime checkpoint between your agent and the tools it calls — GitHub, shell, databases, deploys. It doesn't matter which model or framework you're using. If your agent can call a tool, Relay can protect it. Open source. One command to set up. Would love feedback from anyone shipping agents into production — especially if you've had a close call of your own.

Forum Threads

p/relay-ai

•

4d ago

What's the scariest thing your AI agent has almost done?

Building Relay a runtime security layer for AI agents because my own agent exposed live database credentials with zero warning. No checkpoint, no policy, nothing. Just decided and acted.

Curious if others have had close calls. What's the most dangerous thing an agent tried to do in your setup whether it got stopped or not?

View all

Relay Security

Block dangerous AI agent actions before execution

Block dangerous AI agent actions before execution

Forum Threads

What's the scariest thing your AI agent has almost done?

Forum Threads

What's the scariest thing your AI agent has almost done?