What’s new in AI-assisted development: better integration, new power profiling and hardware debugging skills

What’s new in AI-assisted development: better integration, new power profiling and hardware debugging skills

AI-assisted development is entering a new, more hands-on phase. The new set of features released in September 2026 for the nRF Connect extension for VS Code and the Nordic MCP will make things much simpler for developers, enabling them to boost their embedded development workflows and improve performance and efficiency with AI.

Internal enhacements

The latest updates to Nordic MCP and the nRF Connect extension for VS Code come with several changes under the hood. These changes might not be as eye catching as the new hardware interaction features, but will definitely help you to get better results from your AI assistant.

Better context from Nordic MCP

The first update is a planner feature, that gives your AI assistant a better way of understanding and executing your requests. How is it done? The Nordic MCP now exposes new resources that help the language model break a task into smaller, more manageable steps. By working through a clearer procedure, the AI assistant can complete tasks faster, use fewer tokens, and get more reliable results because the model has a better understanding of the problem.

On top of that, Nordic MCP now qualifies the information sources supplied to your AI assistant. The knowledge base available to supply context to the model is large, but also very diverse, as it includes Nordic TechDocs, DevZone blog posts, DevZone tickets, and more. Sometimes there can be differences or contradictions in the content, not only because of version and product evolution, but also because of human or environmental errors, especially in support tickets, which can be very scenario-specific. In such scenarios, the MCP will now give priority and more importance to the official documentation, while still using alternative sources when there is no relevant information in the documentation, for example.

There is another upgrade in the internals of Nordic MCP that will enable you to access your data in nRF Cloud without needing to configure nRF Cloud MCP, and using the same myNordic SSO credentials. This makes it easier for developers to incorporate real data from the field into the development process and gain a better understanding of the bugs and problems reported in the field while working on them in the lab.

Deeper integration with your development environment

Adding AI assistance to your existing working projects is also getting simpler. The latest version of nRF Connect for VS Code can now share information about your current applications, builds, configurations, and connected devices with your AI assistant of choice. That makes it easier to give an AI boost to the projects you already have under way, or have completed and archived in the past.

The integration is not only about simplicity, as it has an efficiency side too. For example, to analyze a build error, a standalone AI assistant must check the file system, look for the build log files, read them, and only then can it actually start dissecting the problem. The file system search and file retrieval operations might not have a significant cost in terms of tokens, but they can certainly be avoided if the VS Code extension supplies the information directly.

And the icing on the cake: Nordic MCP now exposes skills that let your AI assistant control hardware tools like the Power Profiler Kit II or a SEGGER J-Link probe. This enables many powerful and interesting use cases, helping developers improve their efficiency in key areas like debugging problems and evaluating the power consumption of their designs.

Power profiling with AI

Low power is an obsession at Nordic Semiconductor. Our products are market-leading in this area, and we know it’s one of the main reasons customers choose Nordic parts. We also make tools like the Power Profiler Kit II, an affordable gadget that lets you dive deep into the complex energy consumption dynamics that will affect your product’s battery life.

Although simple, checking power consumption during development phases may be tricky and time-consuming, given the work-in-progress nature of the codebase. To make things worse, your team may be more focused on features and functionality, so small changes in the codebase might silently affect your energy budget. The earlier you spot such things, the better. Nobody likes unpleasant surprises on release dates!

Nordic MCP has a new skill that enables your AI assistant to control a connected PPK2. Instead of watching the screen yourself, you can use AI to capture power measurements over long periods of time, and then look for anomalies, such as extreme values, or compare the results against a reference consumption pattern.

Debugging with a touch of AI

Debugging is still a very manual art. Finding the root cause of a problem can require a lot of expertise and knowledge to understand the problem, know where to search and what to search for, and sometimes also a dose of good luck to find out exactly what’s going on. And of course, it might require several iterations, which means time.

Nordic MCP has a new skill that enables your AI assistant to run GDB sessions and control your debugger probes. This new feature opens up many new possibilities, whether for performing the basic investigation and triage of a bug, gathering insights into how your firmware is performing on real hardware, or letting your agent close the loop for validating a fix. You now have a powerful wingman on your side to help navigate any complicated issue you may encounter.

Conclusion

AI-assisted development is evolving from a helpful advisor to a more active partner in the full embedded development workflow. With richer context, tighter IDE integration, and the ability to interact with real tools and hardware, AI can now do a better job of helping developers become more efficient and productive in their daily tasks. This evolution does not replace engineering judgment; it amplifies it, helping teams move faster, learn more from their devices, and focus their expertise where it matters most.