Milan Mraz
Co-Founder & CEO
Focused on product, growth and bringing Sulcus to the teams building with AI agents.
About Sulcus
Agents need supervision while they run.
Sulcus is building runtime supervision for AI agents: live visibility into what an agent is doing, and the ability to step in. We started with agents that Sulcus runs itself, in Sulcus Cloud.
The framework should define the workflow.
The runtime should make execution accountable.
Why Sulcus
The hard part starts when the agent runs.
Agent frameworks make it easy to define what an agent should do. Once it runs, the questions change: what is it doing right now, which models and tools is it calling, how many tokens has it used, and why did it fail?
Sulcus answers those questions while the run is happening, and gives you a way to stop it. We work with the frameworks developers already use rather than asking them to rebuild around a new abstraction.
Founding team
The people building Sulcus.
Elariz Taghiyev
Co-Founder & CTO
AI engineer and builder focused on agent infrastructure, runtime systems and developer tooling.
Direction
Where Sulcus is today, and where it’s going.
Supervised Cloud runs
LangGraph, CrewAI and OpenAI Agents SDK apps from public Git repositories, with live visibility, token limits and stop control.
Local coding agents
Observing Codex and Claude Code sessions launched through the Sulcus CLI, from the same Runs view.
Beyond Sulcus Cloud
Supervising agents that run on your own infrastructure, configurable run policies and managed secrets. None of these are available yet.
01
Complement, don’t replace.
Work with the frameworks developers choose instead of asking them to rebuild around another agent abstraction.
02
Make boundaries visible.
Be clear about what Sulcus can see and control, and what stays your application’s responsibility.
03
Earn trust through specifics.
Prefer concrete product behavior over broad claims about autonomy, governance, security or scale.