At Luxonis, we believe in building products alongside the engineers and developers who use them. We are thrilled to introduce a new way to collaborate with us: Prototype Access.
To kick off this program, we are launching three brand-new accessories designed to give your M8-equipped OAK devices the ability not just to see the world, but to physically interact with it.
Here is everything you need to know about Prototype Access and the new hardware coming your way.
Prototype Access is a new section of our shop where we stock limited quantities of functional, initial-version devices. It gives you early, ungated access to new Luxonis hardware before it reaches full production.
These products are real and working, but they are still in active development. By participating, you get to evaluate new architectures early, test them in real-world environments, and directly influence the final product through your feedback.
What to expect when you buy a Prototype Access device:
Functional hardware: Devices are actively used internally and by select customers, though some features are still in development.
Evolving specifications: Performance, firmware, mechanical details, or connectors may change based on feedback.
Faster iteration: Expect frequent firmware updates, software improvements, and documentation additions as we learn from real-world use.
Limited availability: Quantities are constrained and not guaranteed long-term.
What Prototype Access is NOT:
It is not final production hardware.
It is not guaranteed to support every intended feature from day one.
It is not intended for customers who require strict long-term supply guarantees, locked specifications, or frozen designs.
Who is this for? If you are an engineer, robotics developer, or R&D team comfortable adapting to early hardware and you want a direct voice in product development, Prototype Access is the fastest way to build with what’s coming next. If you require a fully finalized, production-stable product, we recommend waiting for the general release.
The Launch: Giving OAK the Ability to Act
Vision is only half the equation in spatial AI. A camera can see a problem, but on its own, it cannot fix it.
The OAK4 was designed to be a powerhouse of standalone edge compute. It has the processing capability to run complex neural networks, track objects, and understand 3D space entirely on-device. However, in traditional setups, the camera still has to pass that information to a host PC, which then tells the rest of the system what to do.
We want to remove that bottleneck. By leveraging the M8 I/O connector(available on OAK4 and OAK-D S2 devices) we are launching three accessories that allow the camera to act as the central brain of your system, bypassing the host PC to communicate directly with machines, vehicles, and the cloud. These devices turn your OAK from a passive observer that captures data, to an active orchestrator that can physically interact with the rest of your system and the real world.
The M8 Controller Box is essentially an industrial I/O breakout module. It provides the physical connections, like plug-in screw terminals, needed to wire your camera directly into a factory or automation environment.
Instead of routing decisions through a separate computert o translate the camera’s findings into physical actions, this box equips your OAK with 16 GPIOs, 4 heavy-duty power relays (handling up to 16A/400VAC), RS232 Serial, native CAN, and two USB-A ports. Practically, this means your camera can directly control external lighting, trigger a motor, read a PLC, or switch an actuator the millisecond it detects an anomaly.
If you are building autonomous mobile robots (AMRs), agricultural tech, or heavy machinery, your systems likely run on CAN bus. The M8 CAN Adapter is a compact, plug-and-play module that gives your OAK device a native Linux SocketCAN interface (supporting CAN2.0A and B up to 1M baud).
By connecting directly to a vehicle's internal network via the industry-standard M12 connector, the OAK4 can observe its environment and immediately send steering commands to a tractor or velocity adjustments to a warehouse robot. It brings spatial AI directly to the control systems of heavy machinery, drastically speeding up deployment and eliminating the need for bulky intermediate computers.
Sometimes, running an ethernet cable is the hardest part of a deployment. The M8 WiFi Adapter is built for those environments: sprawling factory floors, smart city intersections, and retail spaces where you need vision but can't easily wire a network.
Because the OAK4 processes all its AI models directly on the edge, it doesn't need to stream heavy, bandwidth-hogging video feeds. This rugged, waterproof adapter allows the camera to sit entirely untethered, running its neural networks locally and efficiently transmitting only the necessary, low-bandwidth metadata (like object counts or safety alerts) back to your central system or Luxonis Hub.
A Note on Prototype Capabilities
Because these three accessories are launching through our Prototype Access program, you may notice the devices look and feel prioritize early functional testing over final retail polish, and their software features are actively rolling out. While the core hardware is fully integrated, certain communication protocols or specific operational modes may still be in development.
We encourage you to review the exact support status of each device before purchasing. Full details, specifications, and current capabilities can be found on their respective product listings:
If you are ready to experiment early and help shape the future of standalone edge AI, we can't wait to see what you build.
