AI is changing the way developers write code, but it is also creating a new problem: there is suddenly a lot more to keep track of. Instead of sitting in one coding environment and writing everything themselves, developers can now work with multiple AI tools and agents at the same time. That means switching between tabs, checking which agent needs attention, sending prompts, approving actions and, sometimes, simply waiting for an AI agent to finish its work.
Logitech believes this has created a need for something that has traditionally been missing from the developer’s desk: a physical control centre for AI-assisted coding. And that is where the company’s recently-unveiled MX Keypad comes in. Designed for developers, the device can be customised to trigger actions, monitor running AI sessions, switch between agents and execute repetitive sequences through a physical set of screen keys.
But Logitech isn’t treating it as a fixed piece of hardware. With AI tools changing rapidly, the company says the MX Keypad has been designed to remain flexible through integrations and customisable plugins. Its partnership with GitHub is also aimed at bringing Copilot into this workflow.
During the launch event held at New Delhi, we spoke to Valeriya Kostyukovskaya, Global Business and Marketing Manager for MX at Logitech. The conversation ranged from what prompted the company to build a dedicated device for developers, how the MX Keypad fits into an AI-assisted coding workflow, and more. Here are excerpts from the conversation.
Also read: Nvidia DLSS 5 will not be limited to 50-series GPUs: All details
Q. What made Logitech decide that developers needed a dedicated piece of hardware in the first place? What problem were you seeing that existing keyboards and mice weren’t solving?
Valeriya Kostyukovskaya: The problem we saw was that agentic AI is fundamentally changing the coding workflow. We saw a shift that was happening extremely fast, and the number of AI tools was also growing and changing very rapidly. As a result, we started seeing friction between different multitasking activities and switching between different agents and tabs. I think it is not only about the agentic workflow, but the whole layer cake of different tasks that developers do.
However, the agentic workflow is where we saw the potential and the gap for a tactile, physical control centre. It not only allows you to trigger actions, but also lets you monitor what is happening. So, for us, it has always been about this dual purpose of control and monitoring.
Q. What are some of the most common actions you expect developers to assign to the MX Keypad?
Valeriya Kostyukovskaya: I think the most common use is assigning running sessions. Developers can take whichever tools they use, whether it is Codex or Cloud Code, and assign some of their running sessions so they can monitor their state. They can also assign some very basic controls for their sessions, such as yes or no responses or predetermined prompts that they usually send to the chat. Voice control of the agent is another use case.
Those are probably the most-used actions. I also see this reflected in the content being produced by different influencers, because these are the actions they talk about using the most.
Q. AI-assisted coding is changing the role of developers. How has this shift influenced what you wanted the MX Keypad to do?
Valeriya Kostyukovskaya: For us, what’s been a challenge is that we’ve been preparing the product for launch for a while. I’ll give you a funny example. We filmed the main (promo) videos back in spring, and when we were filming them, we had a certain set of tools as our hero tools. Now, that has already changed from the tools that we were filming. So basically, it is all changing so fast that for us, it’s important to react and bring the needed integrations.
We also want to facilitate our community to build those plugins. Because in another six months, the tools might change completely. We don’t know. New AI tools are popping up all the time.
So we definitely want to stay agile and bring new integrations. I think the fact that the product is so versatile, it has the screen keys, it’s a simple form factor and there are so many things you can customise, means it will be able to adapt fast.
Q. Can you share an example of an AI-assisted coding workflow that becomes easier or faster with the MX Keypad?
Valeriya Kostyukovskaya: Repetitive routines are one of the biggest use cases. If a developer is debugging or working on a pull request, for example, they can create a single button that executes a sequence of steps. This can include predetermined text as well, so if they need to enter a description, that can be built into the automation.
The other major use case is monitoring and switching between agents. Even simply being able to switch between agents can make a big difference. You do not necessarily want to interact with the agents using the keypad, but having visibility into which agent needs your attention is extremely useful. For example, the keypad can blink with a notification mark, and when you press it, that particular agent comes up.
That is important because otherwise developers would need to have multiple screens set up for each agent.
When we spoke to developers about this, one very senior developer told me that the waiting time while an agent is running was a real challenge. There are periods when the developer is not doing anything, so previously they would just sit and stare at the agent. Now, they can hide the menu, move on to other tasks and still monitor the agent in the background.
Q. Did feedback from developers for previous Logitech products influence the design and functions included in the MX Keypad?
Valeriya Kostyukovskaya: Yes. The integrations and use cases, as well as the way we built the plugins, were fully influenced by the community.
We have this MX Collective, which is a new Discord community full of MX users. It is a really vibrant community, and the users are very engaged and vocal. We also created a separate developer community that was more focused on developers.
We had a hackathon hosted in Lausanne, I think last January or last winter, so there was a lot of engagement from the community in shaping the use cases and integrations.
GitHub has also played a big part. We worked with GitHub for almost a year, and they helped develop the integration and tested the product. They believe in the product, and for us, that is a big win.
Q. How does the GitHub Copilot integration work in practice, and what can developers do with the MX Keypad that would otherwise require multiple steps?
Valeriya Kostyukovskaya: The Copilot integration is similar to what you may have seen in the Cloud Code videos from creators. You also have the agents and sessions, and you can control those sessions through the standalone Copilot app.
That integration is still being worked on by the GitHub team, but it will be published, or it may already be published.
We have one more integration inside Visual Studio Code because Copilot is a native part of Visual Studio Code. There, we have customised some of the most-used commands that developers use inside VS Code to trigger Copilot.
You can trigger Copilot either through the chat on the side, or Copilot can interact directly with your code through what is called inline Copilot. We integrate with both.
You can also program different prompts that you want to trigger with the chat inside Options Plus, which is quite useful as well.