Progress Software has released a new version of Telerik and Kendo UI that pushes AI past developer assistance. The updates aim to let developers build applications where not only people interact but also AI agents that can use interfaces, process documents, query enterprise data, and take actions. Notably, it adds support for WebMCP, meant to give agents a more structured option than visual recognition and screen scraping.
Telerik and Kendo UI in 20 seconds
- Progress upgrades Telerik and Kendo UI with tools for building applications ready for AI agents.
- WebMCP lets applications expose interface functions in a structured way so agents can use them.
- Fiddler now lets programming assistants analyze network traffic during debugging.
- Document libraries add tools to process Excel files, extract data, and generate files through agents.
- New features also include RAG, WinForms modernization, and AI-driven interface generation.
The release reflects a shift starting to matter in software development. Until now, much of AI integration focused on helping programmers write code. The next phase is preparing applications so agents can work with them directly.
That’s a technical challenge. Traditional graphical interfaces are built for people: buttons, menus, forms, and visual elements users read easily. An agent has to discover what it can do in an application, understand what each action means, and carry it out reliably.
Progress wants to cover both sides: using AI during development and building interfaces that other AI systems can use later.
WebMCP aims to keep agents from having to read screens
One of the more interesting additions is WebMCP support in Progress’s interface tools.
Current agents can handle web pages in several ways. Some use vision models to interpret what’s on screen; others inspect the Document Object Model (DOM), automate browser actions, or use screen scraping.
These work, but they can be fragile.
A moved button, a layout change, or a website update can break the automation.
WebMCP proposes a different route: letting a web application expose certain functions in a structured way that agents can discover and use.
An example helps.
An AI that needs to book an appointment might try to open a page, visually find the calendar, pick a date, spot a free slot, and click the right button.
A dedicated agent-friendly interface could instead just tell it which operations are available, the parameters they need, and what each returns.
The person would still see and use the usual graphical interface, but the agent would have a software-optimized layer.
Progress says this can cut reliance on visual automation and make interactions with applications more reliable for agents.
Fiddler will let AI assistants see network traffic
AI also plugs directly into the tools developers use.
Progress introduces Agent Skills for Fiddler, its well-known platform for debugging and analyzing HTTP traffic.
The idea is to give programming assistants the context a developer usually watches when troubleshooting.
An agent could analyze network traffic, spot abnormal behavior, and suggest possible causes and fixes.
That addresses a limit of current programming assistants.
A model can analyze code within the context it gets, but errors aren’t always in the file on screen. They might be in an HTTP request, an API returning unexpected data, a wrong header, or a communication issue between services.
Giving it access to that information widens the debugging context.
Agents will also work directly with documents
Another part of the update touches the Telerik Document Processing Libraries (DPL).
Progress adds tools so agents can perform document operations inside the applications themselves.
These include extracting structured data, analyzing Excel spreadsheets, converting formats, and generating documents.
The company stresses these operations can be built into existing applications without necessarily relying on external AI services for document processing.
That’s especially useful in enterprise settings.
An integrated agent could take in spreadsheets, pull specific information, structure the data, and generate new documents using the available libraries.
So AI moves from just discussing documents to actively working with them through specialized tools.
RAG now integrates directly with .NET applications
Progress is also adding Retrieval-Augmented Generation (RAG) capabilities for .NET developers.
RAG connects a large language model with external information sources outside its original training data.
In a business context, that could mean a system consulting internal manuals, contracts, technical documentation, or data repositories before it answers.
The new version lets .NET applications integrate with Progress’s RAG-as-a-Service offering, so models can work over unstructured enterprise data.
Compared to a general-purpose chatbot, the key is the context you supply.
An application can feed the model only the relevant information, so responses draw on the organization’s documentation.
This doesn’t remove issues like inaccurate model responses or guarantee correctness, but it gives a better architecture for handling confidential or specialized knowledge.
AI will also help modernize WinForms applications
The update also tackles a less flashy but very common problem: legacy software.
Progress adds automation tools to convert old interfaces to Telerik UI for WinForms.
The process uses smart control mapping, code analysis, and Model Context Protocol (MCP)-based automation to help with the transformation.
The goal isn’t only generating new interfaces from scratch but cutting the effort to modernize existing applications.
WinForms is still common in many enterprise systems built over decades. Migrating those apps takes time because they hold accumulated logic and components.
AI is starting to help with this less visible work of maintaining and updating legacy software.
From using AI for programming to programming for AI
The Telerik and Kendo UI update stands out for a reason beyond its features.
In recent years, tools like GitHub Copilot and various programming agents have brought AI into the developer’s environment. Models write code, analyze repositories, run tests, and suggest fixes.
Now the reverse is emerging: developers have to build software knowing that other intelligent software will also be a user.
That can shape how applications are designed.
A user-only interface has to be visually clear. An agent-enabled app also has to clearly describe the available operations, the required inputs, and the limits.
Standards and protocols like MCP and WebMCP point that way.
Progress talks about agent-ready applications, though how much this catches on outside experimental projects remains to be seen. Still, the idea fits the evolution of agents: as they get able to complete whole tasks, applications have to provide reliable ways to interact.
The future of AI in development may involve not just programmers using agents to work faster, but rethinking how software is built so humans and agents can use the same applications in different ways.
Frequently Asked Questions
What new features are added to Telerik and Kendo UI?
The update adds AI-driven interface generation, support for agent-ready applications, document processing, assisted debugging, RAG integration, and tools for modernizing WinForms applications.
What does WebMCP offer AI agents?
It lets applications expose certain functions in a structured way so agents can interact with them, cutting reliance on techniques like screen scraping.
What changes in Fiddler?
Progress adds Agent Skills that give programming assistants visibility into network traffic to help catch issues during debugging.
Do the new tools include RAG?
Yes. Progress lets .NET applications connect to its RAG-as-a-Service platform, so models can work with documents and other unstructured business data.
via: investors.progress
