Hey Product Hunt 👋 I built CTRLRun because AI agents are getting better at deciding what to do, but the risky part starts when they can actually change real systems. A simple example: an agent issues a refund, the provider processes it, but the response times out. The agent thinks it failed and retries. Now the same action may happen twice. CTRLRun sits in front of consequential actions and adds approvals, policy checks, duplicate-action protection, ambiguous outcome handling, and execution receipts. It is open source and works with Python, LangGraph, OpenAI Agents SDK, and MCP. Still early, so I would really value feedback from people building agents that can modify real systems. GitHub: https://github.com/CTRLRun/ctrlrun If CTRLRun looks useful, a GitHub star ⭐️ would genuinely help the project get discovered. And if you are building agentic systems, I would love to hear what failure cases you are dealing with.

CTRLRun

Execution safety for AI agents that take real actions

Execution safety for AI agents that take real actions