The development environment for robotics.
Describe your robot, wire its behavior from modules. Compile it to a native binary for your hardware. Run ROS as a module for deeper debugging metrics.
Shinro is a development environment for robotics. Describe your robot, wire its behavior from visual modules, and compile it to one native binary that runs on the cake kernel across heterogeneous hardware — Jetson, ESP32, and custom boards. Run ROS as a module for debugging and interop. Shinro Studio is source-available under fair-code terms and free to use; community modules carry their authors' licenses; the cake kernel is commercial, distributed as-is.
Teams rebuild the same plumbing instead of shipping what makes their robot unique.
Every robotics team writes its own drivers, deployment pipelines, debug tooling, and hardware abstraction. Vendor SDKs lock you to one platform. Internal stacks take years to develop. Physical AI requires fast iteration
Too much robotics engineering time disappears into non-differentiating infrastructure.
Different robots. The same plumbing, rebuilt from scratch, every time.
Two shifts created the gap.
Five years ago a team picked one platform. Today every team spans compute targets: Jetson for perception, ESP32 for low-level control, custom boards for accelerators. Existing tooling assumes one target.
Robot behavior is composed from learned policies, perception models, and controllers. Every team builds the integration plumbing itself. The development environment for AI-native robotics does not exist yet.
every tech wave produced the company that owned the workbench. games got Unreal. code got GitHub. data got Databricks. robotics still rebuilds the workbench project by project.
Same stack. New foundation.
Shinro sits between robotics frameworks and hardware targets: one consistent way to compose, build, deploy, and observe robot software across runtimes. Run your stack with less runtime plumbing.
Build once, run on anything.
A traditional layered stack. Each layer talks to its neighbors, and a change means rebuild and restart.
One thin layer above hardware. Connecting any module to every layer.
One environment: describe the machine, compose behavior, simulate, deploy native, swap live.
See everything.
Drag, drop, and link blocks to map hardware and orchestrate behavior.
Real-time physics for pre-hardware validation, live telemetry after deployment.
Pick your target device and settings, then swap live modules at low latency.
Where composition happens.
Components find each other and exchange messages while the robot runs.
Wiring is resolved while the robot runs. Every hop crosses the bus.
The toolchain reads your module descriptors and generates the inter-process plumbing into one binary.
generated plumbing: shared memory, pipes, concurrency
target hardware: any architecture
Deploying updates a module on the kernel, at runtime. The rest keeps running.
Shinro moves the hard part from runtime to compile time.
The developer loop.
Compose in Shinro Studio, or reuse existing modules.
The kernel toolchain cross-compiles your modules for the target architectures.
Ship to the device, or hot-swap a module at runtime without stopping the robot.
Telemetry streams back from the robot.
Develop, build, deploy, observe. Then iterate.
Three components.
The AI-native visual programming and development platform.
Community-published modules, ready to drop into your build.
A low-latency, cross-platform executable. Distributed as-is, deployable to any architecture.
Develop in Studio · run on cake · share what you build.
Live module updates without downtime.
Add hardware capabilities or replace software behaviour without stopping the task.
A digital twin of a desktop robot arm continuously picks a block and places it on a target marker. Drag the block anywhere and the arm replans without stopping.
Mount modules while it runs: a gripper enables carrying, a camera adds live perception, force control softens contact, a mobile base extends reach, and swapping the planner changes trajectory quality mid-motion.
No restart at any point.
What is open. What is licensed.
Source in the open, free to use. Fair-code terms.
Each module carries its author's license. Open source encouraged.
Commercially copyrighted, distributed as-is.
Open tools to build with, a dependable kernel to run on.