I work on systems that stay up, changes that stay safe, and jobs that should not be done by hand twice. Most of that work is in Go, on Kubernetes, and in the path from code to production.

DMR means Direct, Mediate, Regulate. It is a Go framework for AI agents.

• Direct — you set the goal and the limits.
• Mediate — the agent calls tools and carries the work through.
• Regulate — policy, approval, and an audit log keep it in check.

In DMR we trust, but verify.

DMR itself is closed source. The library behind it, dmr-devkit, is open source. You can embed an agent, tools, workflows, and a record of each run in your own app.

I treat an agent as a teammate you can assign work to. It needs clear permissions, a record of what it did, and a person on the step that matters. The bar is the same one I use in SRE: you can watch it, check it, and roll it back.

• SRE: stability, change risk, automation, and turning repeated work into a platform
• Stack: Go, Python, Java, Kubernetes
• Open source: dmr-devkit and my other repositories
• Talk to me about: SRE, platform engineering, and agents you can audit
• Email: seanly.me@syscube.dev

dmr-devkit — a Go library for building agents you can embed: the loop, tools, workflows, and an audit log of each run.