Your agent needed a platform. Now it has one.

A year ago, developers were arguing about which agent was best for vibe coding for their weekend app. Today, every developer I talk to has a preferred coding agent, and it’s writing more code than they do. The question I keep hearing is no longer which agent to use, but why a working prototype still takes months to ship reliably into production.

The gap in moving from generated code to production execution is real, and today’s announcements address it head-on.

The gap isn’t one of the intelligence or speed of the coding agent. The industry has spent the last two years pouring effort into better prompts, model context protocols, evaluators, skills, harnesses, and sandboxes. But this gap persists. The bottleneck isn’t about agent capabilities, but the platform that runs beneath it.

Providing a production-grade platform is the work that UiPath has been doing for a decade—a platform that enables robots, humans, and now agents to come together in a scalable and secure way. Today, that platform is now native to how you build agents in the enterprise.

UiPath for Coding Agents, available today

Starting today, any coding agent—the one in your terminal, the one in your IDE, or the one you wrote yourself—can natively operate at every layer of the UiPath Platform™. This isn’t a thin CLI wrapper or a code generator; the CLI provides full-platform access. You can plan an end-to-end process, build a UiPath workflow, perform a reliable 360-degree test, detect and debug a failure, publish it into a scalable production environment, and govern it with enterprise-grade rigor.

You can now use natural language to describe what you need from the comfort of your IDE and existing inner-loop, and your coding agent now knows how to harness the UiPath Platform.

This is what this means for our customers already using a coding agent:

• We meet you where you are. You don’t need to adopt a new tool or become a UiPath expert. You continue to build using your IDE or agent of choice, and ship directly into UiPath using the repos, continuous integration practices, and coding conventions that you use today. And our integration connectors handle production systems connectivity in the process orchestration layer.

• Faster, and safer. With UiPath coding agents, you are not writing code from scratch. You are configuring a platform we have battle-tested for a decade against the most rigorous demands of our customers. That gives you time to value. Critically, it also gives you the benefits of a decade of hardening: validation and packaging without friction, durable execution that recovers from failure, end-to-end outcomes that hold up in production.

• Your automations stay maintainable. The newly released UiPath skills and CLI steer your coding agent to follow explicit platform conventions, dependency alignment, and reusable targeting patterns by default. Your agent takes advantage of UiPath expertise to build and configure a production-ready automated process that is designed to run on our governed enterprise platform.

• Governance is built-in, not bolted-on. Automations generated by coding agents automatically enter a governed environment where policy enforcement, auditability, and runtime controls are standard.  Every action is traceable, auditable, and recoverable. It’s automation development at AI speed, without the hidden compliance risk.

Our approach is purposely agent-agnostic. Bring the coding agent you already use, select the ones you want the skills added to, and you’re up and running. Your choice of agent should never lock you into a platform, and your platform should never dictate your choice of agent.

UiPath Founder and CEO Daniel Dines wrote a blog post earlier this week about UiPath for Coding Agents. Read his blog post.

An AI-native design canvas, coming soon

At DevCon, we also previewed an upcoming design surface for AI-native builders: Maestro Flow.

Maestro Flow is a new way of building on the UiPath Maestro™ layer we introduced last year. This new design canvas provides developers with:

• One artifact for every actor. APIs, agents, robots, and humans are present in the same process definition, with handoffs and connections defined and captured in a single canvas. Building your process using a handful of interdependent, separately versioned workflows is finally a thing of the past.

• Build where you are. Whether you build in the browser, in VS Code, or in the coding agent terminal, the Maestro Flow meets you where you are…and BPMN (Business Process Model and Notation) experience is not required.

• Durable at scale. Because your processes run on the same Maestro layer, execution is fault-tolerant by default. You don’t need special configurations to maintain state across restarts, deploys, or long-running approvals.

If you’ve spent any time trying to glue together a process orchestrator, an agent framework, a workflow engine, and an observability stack to get something production-shaped, you understand why the unified canvas is a big deal (and why people who haven’t built one in production tend to underestimate it).

From the start, we’ve been building Maestro Flow to be the all-in-one visual canvas to build, verify, and ship the work your agents are doing. And this new design canvas will ship with full coding agent support, meaning that you can describe what you want to orchestrate—in plain language, in code, or some mix—and quickly get a complete, runnable, production-ready process you can inspect, edit, and ship.

Maestro Case to orchestrate the work that won’t sit still

Maestro BPMN excels at structured processes such as invoice approvals and deployment pipelines. They have a shape and a design that may be complex but defined. But what about the work that doesn’t sit still?

Maestro Case manages processes that are dynamic in nature and where context guides the next step. For example, consider an insurance claim with three claimants and a contested cause of loss—the business’s focus is on the case and not the workflow. Other examples include compliance investigations, loan processing, and customer escalations that pull in three or more teams.

For these types of processes, Maestro Case pursues a defined goal where the path varies per instance:

• Case agents govern, they observe and act across the full case lifecycle, handling assignment, escalation, summarization, and resolution recommendations as conditions change

• Stage agents manage execution and do the structured work inside each stage, whether that's intake, verification, fraud review, or settlement

Together, they replace the manual patchwork that typically lives in someone's inbox or a shared spreadsheet. The result is governed decision making that can execute at scale.

Maestro BPMN is for flow complexity: the graph may be large, but given an instance's state, the next step is determinable. Maestro Case is for context complexity: the next step is a judgment over accumulated, open-ended context that you couldn't fully enumerate at design time.

Maestro Case is now available in private preview.

Data Fabric provides live enterprise context for every agent

Federated entities in UiPath Data Fabric can now give your agents visibility into live data from wherever it lives, in real time, without duplicating the underlying data from Snowflake, Databricks, or any other data system.

Every AI agent is only as good as the context it gets, meaning data is the biggest thing holding enterprise agents back. And the enterprise context for these enterprise-grade processes is often not in one place - it spans multiple platforms and systems of record – and exposing that context to an agent means copying it, moving it, or syncing it. The result is often stale data that can create governance gaps and decisions made on yesterday's numbers.

Zero-copy connectivity to your major enterprise data platforms and systems changes that. Your agent gets the relevant context at the moment it needs it.

And once modeled in Data Fabric, you can reuse that context across your managed agents, automations, apps, and workflows. The result is a cleaner architecture that works off fresh data and complete context, yielding decisions you can defend and processes that don’t drift between systems of record.

Testing that keeps pace with code that writes itself

Software testing has been a core investment area for UiPath for years. Gartner, Forrester, and IDC have all named UiPath Test Cloud a leader across their respective evaluations.

Today, we demonstrated how UiPath Delegate—our new computer-use agent that can execute enterprise work across apps, files, and browsers—can empower and accelerate the work of testing teams. As coding agents generate more code at higher velocity, testing must scale with them. Customers leveraging our testing solution today are seeing six times faster release cycles, shipping in days instead of weeks, with roughly $4 million in annual savings from reduced testing overhead. UiPath Test Cloud—now joined by Delegate—is how we keep software testing moving at the speed of change.

• Autonomous test case execution with UiPath Autopilot™ runs your manual test cases end-to-end without a human at the keyboard. It works across the web with Playwright, mobile with Appium, and it can test in the enterprise systems where the real work lives—SAP, Oracle, Epic, and more.

• Exploratory testing at scale. Hand Autopilot a goal, and it maps what it finds and reports back what’s worth a human’s attention.

• Eliminate tedious tasks for testing teams. Roughly 70% of a tester’s time goes to administrative overhead—test management updates, Jira hygiene, status reporting, build coordination.

Testing teams can use Delegate to drive down this drag. We will share more information in the coming months on Delegate and how testing teams can use it to transform the way they work; watch this space.

Build at agent speed. Orchestrate at enterprise scale.

The hard part of AI agents was never the demo - it was getting them to run reliably in production with the controls and governance a CISO can sign off on. Today we shipped the next chapter of the work that the UiPath Platform is known for. The rest is yours to build.

If you weren't able to join us in person at DevCon, we will be sharing the most popular sessions on demand starting May 22. Register now to get the DevCon Best Bits delivered right to your inbox.