What is Telekinesis?
Telekinesis SDK is a large-scale skill library for Physical AI. It provides a unified, composable set of algorithms for robotics and computer vision, covering:
- 3D perception including detection, registration, filtering and clustering
- 2D perception including image processing, detection, and segmentation
- Synthetic data generation
- Model training tools
- Motion planning, kinematics, and control
- Physical AI agents
- Vision-Language Models (VLMs)
The SDK is designed to reduce integration overhead by providing consistent abstractions and interfaces across these capabilities, allowing roboticists and computer vision engineers to focus on system design and application logic rather than stitching together fragmented tools.
Our thesis is to build a complete all-in-one skill library that can be orchestrated by Physical AI agents to perform any task.
A large-scale skill library orchestrated by Physical AI agents
Flow Overview
- A user expresses a task at a high level (for example, a CNC machine loading request).
- A Physical AI agent interprets the intent and reasons over available skills.
- The agent constructs a high-level plan as a sequence of reusable skills.
- Skills are executed safely on the robot using validated motion and control primitives.
How Does Telekinesis SDK Work?
Telekinesis provides a single, consistent interface to compose all perception, planning, control, and AI skills into complete robotic systems.
Using Telekinesis SDK
The Core Problem
Robotics software development is fragmented. There are multiple players: teams building models, groups developing applications, open source projects, and specialists focused on perception, planning, or control. Each has its own APIs, conventions, and language.
Building even a single application often requires stitching together pieces from multiple sources: classical robotics libraries, computer vision modules, deep learning models, and foundational AI models like LLMs and VLMs. This slows development, introduces integration errors, and makes it difficult to maintain coherent, industry-grade systems.
With the rise of Physical AI and AI-driven agents, this fragmentation is becoming more acute. Developers face a growing gap between classical robotics workflows and modern AI capabilities, making it hard to experiment, iterate, and ship solutions efficiently.
How Telekinesis is Different?
Telekinesis unifies classical robotics, computer vision, deep learning, and foundational models in a single SDK. It’s not just another robotics framework: it’s a modular skill library that lets you combine perception, planning, control, and AI in a consistent, predictable way.
Every skill is composable, ready-to-use, and designed for experimentation or production, so you can build complete applications without hunting for libraries or writing integration code from scratch.
Who Is Telekinesis SDK For?
Telekinesis is built for engineers and teams working at the intersection of robotics, computer vision, and Physical AI who need:
- A consistent, unified SDK across robotics, computer vision and AI skills
- Access to a broad set of algorithms to build complete, industry-grade applications
- The ability to experiment with modern AI models alongside classical robotics
- Reduced integration friction and faster development cycles
The Telekinesis Skills
The Telekinesis SDK provides a single Python interface to access hundreds of modular, composable skills hosted on the cloud. These skills cover everything from perception and planning to control and AI agents, allowing you to build complete robotics and Physical AI applications without worrying about integration.
For enterprise deployments, the SDK can also be run on-premise, giving teams full control over data and computation while maintaining access to the same unified set of skills.
This architecture ensures that developers can:
- Quickly prototype and experiment with modern robotics and AI workflows
- Combine and reuse skills flexibly across different applications
- Scale from individual projects to enterprise-grade systems without changing APIs
A complete guide on Skills is available on Skill Concepts.
Where to Start?
Ready to explore Physical AI skills to build your robotics application?