← sources

How are people currently managing AWS access for AI agents? I'm looking into different approaches for securely handling permissions and credentials when agents need to interact with AWS services.

source: reddit (reddit:devops)published 2026-09-19 19:43ingested 2026-09-20 02:36

retained text

I'm currently looking into how to securely manage AWS access for AI agents like Claude Code within our organization. 

Right now, our developers have a role that grants them some write access to specific services in our dev and nonprod environments. However, I really don't want them using this same role when working with AI agents, as I'd like to limit the agents' blast radius. 

I'm trying to figure out the best way to restrict the AI agent's AWS access to read-only, even if the developer using it has broader permissions. How are other teams handling this? Are you creating separate, read-only roles specifically for AI agents, or is there a better way to enforce this boundary?

If a separate readonly role is being used, how do you make sure AI agents would force the readonly role? 

I want something to work with Claude code, Codex, Opencode, cursor or any other AI tool out there.

document fetched 2026-09-20 00:24 · canonical https://reddit.com/r/devops/comments/1wkvk6c/how_are_people_currently_managing_aws_access_for