Requirements, CAD, PCBs, simulation, training and firmware in one versioned workspace, with agents that design alongside you.
Sign in with Google. A personal workspace is created on first sign-in.
Built in AccelerAI
A small wheel-legged biped. Each leg pitches at the hip, swings out on an abduction joint, turns at the thigh and bends at the knee, and ends in a driven wheel, so it can roll, balance and step. The body carries the battery low and the unchanged Vertex Mini PCB above it, with an LED strip behind the front window.
Every part, the assembly and its joint limits, the board, the MuJoCo model and the locomotion policy live in one AccelerAI project. What you see above is that CAD.
One workspace
Agents work through the same tools you do: they read the project, model parts, check fits and clearances, route boards and start training runs, and every change lands as a commit you can review.
Capture what the robot must do and track the work on a board the agents read and update too.
Parametric Part Studios, sketches with constraints, and assemblies whose mates become the robot's joints.
Boards as code with tscircuit, checked, routed and rendered in 3D, sharing one contract with the CAD.
Export to MJCF or URDF, simulate in MuJoCo in the browser and train policies with reinforcement learning.
Run FEA on printed parts, then take the trained policy all the way to the microcontroller.
Every change is a commit. Try wheels instead of feet on a variant, compare, and merge what works.
Open the Vertex Mini example, or begin from a blank project.