CtrlAI

Maaz, Sid Fan, and Gianmarco Carrieri looked at the AI agent explosion and said "wait, who's actually watching what these things are doing?" And honestly, that's a terrifying question. AI agents are out there making tool calls, executing actions, doing things—probably on your dime—and most companies have zero visibility into any of it. So they built CtrlAI, a transparent HTTP proxy that sits between your agent SDK and the LLM provider and evaluates every single tool call against configurable guardrail rules. Zero code changes needed. You just point your agents through CtrlAI and suddenly you've got governance.

Here's what makes this important: AI agents can call tools. Those tools can delete data, send emails, make payments, access sensitive systems. Without guardrails, you're basically handing an toddler the keys to a Ferrari and hoping for the best. CtrlAI intercepts every response, checks tool calls against rules you define, blocks the dangerous ones, audits everything, and includes a kill switch so you can shut the whole thing down instantly if things go sideways. It's open source, it's free, and it works with whatever agent framework you're using. As AI agents get more autonomous and more connected to real systems, someone has to be the adult in the room. CtrlAI just volunteered.