As coding agents become the default way software gets built, Interfaces lets any team member generate a custom labeling or evaluation interface, with Label Studio Enterprise providing the quality control, governance, and scale underneath.
San Francisco June 10, 2026
San Francisco, CA. June 10, 2026. HumanSignal today announced Interfaces in Label Studio Enterprise, which lets teams build any labeling or evaluation interface by prompting an AI agent. The agent generates the interface as production-ready code, and Label Studio Enterprise runs everything beneath it. The launch positions Label Studio as the durable infrastructure layer beneath agent-driven development, rather than another application bounded by fixed templates. With Interfaces, teams can:
Coding agents have moved quickly from novelty to the default way teams write software. The harder problem has always been everything beneath the interface: the quality control, reviewer workflows, access control, audit trails, and compliance that production AI data operations depend on. Interfaces let teams move at agentic speed on the frontend while Label Studio Enterprise handles that operational backbone, so a fast prototype becomes a system a team can actually run at scale.
That backbone is also what separates Interfaces from the wave of disposable, single-purpose tools an agent can now generate in minutes. The interface is easy to create; the governance, quality, and durability around it are the hard part, and they are what make the work viable in production.
Until now, the kinds of work a team could support were bounded by a fixed library of templates. Interfaces remove that limit. Because the interface is generated as real code, teams can build for nearly any data type or workflow they can describe, from document and time-series review to robotics episodes and agent evaluation. That extends Label Studio's reach across a far wider range of AI builders and use cases, including modalities that did not have a home on the platform before.
"Coding agents are becoming a native part of how teams build software, and we want that same shift in the data layer," said Michael Malyuk, CEO of HumanSignal. "With Interfaces, a product or program manager can build a sophisticated labeling or evaluation workflow just by describing what they need, with no engineering team required. The people closest to the data can finally build the interfaces and workflows they need themselves, on infrastructure they don't have to think about."
Early adopters report sharp efficiency gains. One enterprise customer rebuilt its evaluation interface on Interfaces and cut clicks and keystrokes for annotation tasks by 90 percent. Another said the approach unblocks non-technical teams so they can keep pace with demand.
Teams can use the built-in agent, which runs in a sandbox isolated from production data, or bring their own, including Claude Code, Codex, Cursor, or a local model for on-premise work. Either path includes version control, access control, and compliant infrastructure by default.
Interfaces are available in beta on Label Studio Enterprise. Teams can request a walkthrough on a specific use case at https://humansignal.com/contact-sales/.