Dimensional is an agentic operating system for controlling physical robots and building multi-agent systems that operate in physical space.
Dimensional addresses the fragmentation in robotics by providing a unified Python-based SDK that works across different hardware platforms—humanoids, quadrupeds, drones, and arms—without requiring ROS. The tool enables natural language control of robots through an agent-native architecture where agents run as native modules and can subscribe to embedded streams from perception systems like cameras and lidar, as well as control loops and motor drivers. This approach allows developers to write robot applications entirely in Python and deploy them across diverse hardware with a single codebase.
Dimensional suits teams building generalist robotics applications who want to avoid the complexity of ROS integration and prefer a modern Python-first development experience. The platform is particularly valuable for projects requiring multi-agent coordination, natural language command interfaces, and seamless integration of perception and spatial reasoning. The tool includes capabilities for SLAM-based navigation, dynamic obstacle avoidance, route planning, autonomous exploration, object detection, 3D projections, vision language models, audio processing, and spatio-temporal retrieval-augmented generation for spatial memory.
The project is in pre-release beta. Development activity shows ongoing work across core infrastructure, with regular commits addressing agent systems, hardware integration, and perception capabilities. The maintainers are actively expanding hardware support and refining the agent control architecture based on demonstrated use cases.
