Is OpenDevin safe for production codebase refactoring?

All answers to this question.

Safety is actually the core focus of the OpenDevin architecture. Every agent run happens inside a strictly isolated Docker container. The agent can only see the files you explicitly mount into the /workspace directory. It can't touch your host system's root or environment variables unless you misconfigure it. That said, I’ve seen OpenDevin get a bit "enthusiastic" and rewrite files it wasn't supposed to. Always use it on a clean git branch and review the diffs. It’s a great tool for refactoring, but it still needs a human "senior engineer" to hit the merge button.

   Answered 2025-03-05 by Melissa Foster

Can we implement a "Manual Approval" step for every terminal command OpenDevin tries to run?

   Answered 2025-04-15 by Gregory Scott

• Yes, Gregory. There is an interactive mode in the OpenDevin UI where the agent pauses and asks for permission before executing shell commands. This is highly recommended for security tasks. It prevents the OpenDevin agent from doing anything destructive, like an accidental rm -rf, without you literally clicking a "Confirm" button in the web interface.

     Commented 2025-04-22 by Eric Bennett

The sandboxing is solid. I’ve tried to "jailbreak" it from the inside and it’s very well contained within its Docker shell.

   Answered 2025-05-10 by Laura Bennett

• Good to hear, Laura! That level of containment is what makes OpenDevin a viable choice for corporate environments with strict security policies.

     Commented 2025-05-15 by Austin Reynolds