Telekinesis Agentic Skill Library

Appearance

Telekinesis Agentic Skill LibraryThe Intelligence Layer for Physical AI

Quickstart
Create API Key
Github

Cornea: Image Segmentation Skills

Segment images using classical or state-of-the-art deep learning models
Status: Released

Retina: Object Detection Skills

Detect objects in images using foundation models and classical algorithms
Status: Released

Pupil: Image Processing Skills

Perform image processing tasks and low-level vision algorithms
Status: Released

Vitreous: Point Cloud Processing Skills

Process point clouds and build flexible 6D pose estimation pipelines
Status: Released

Illusion: Synthetic Dataset Skills

Generate realistic datasets to train and validate vision and perception models
Status: Planned — Apr 2026

Iris: Model Training Skills

Train, fine tune, and deploy state-of-the-art deep learning models
Status: Planned — Apr 2026

Neuroplan: Motion Planning Skills

Plan robot motion using kinematics, trajectory optimization, and collision-aware planning
Status: Planned — Apr 2026

RLBotics: Reinforcement Learning Skills

Train and deploy reinforcement learning agents for sim-to-sim and sim-to-real robotics
Status: Planned — Apr 2026

Tzara: Physical AI Agents

Orchestrate perception, planning and control using LLM-powered Physical AI agents
Status: Planned — Apr 2026

Telekinesis Agentic Skill Library for Computer Vision, Robotics and Physical AI applications

Telekinesis Agentic Skill Library for Physical AI ​

The Telekinesis Agentic Skill Library provides the intelligence layer for robots. We unify the entire universe of robotics into a single, production-grade autonomy stack: integrating Robot Foundation Models (RFMs), Physical AI Agents, skills, data pipelines, and real-time communication for industrial-grade robot deployment.

The Telekinesis ecosystem is built as a 5-layer architecture covering the entire lifecycle of robotics intelligence:

  1. Industrial Applications: Industry-grade applications ready for deployment across manufacturing, logistics, robotics research, and more.
  2. Physical AI Agents: High-level reasoning agents powered by Vision-Language Models (VLMs) and Large Language Models (LLMs), capable of interpreting natural language and orchestrating complex, multi-step tasks.
  3. Skills: A repository of production-grade, composable atomic actions spanning robotic perception, motion, hardware communication, and other core capabilities.
  4. Data Engine: A high-throughput pipeline for collecting, labeling, and fine-tuning robot trajectories, enabling continuous learning and adaptation.
  5. BabyROS Middleware: Ultra-low-latency, ROS-like communication enabling seamless data exchange between nodes on edge and physical systems.

Installation ​

Get started immediately with the Telekinesis Agentic Skill Library (Python 3.11 or 3.12):

python
pip install telekinesis-ai

💡 Information

A free API_KEY is required. Create one at platform.telekinesis.ai. See the Quickstart for more details.

Industrial Applications ​

Telekinesis Agentic Skill Library helps you build real-world robotics and Physical AI applications for industries such as manufacturing, automotive, aerospace and others. We present some use cases in manufacturing which the Telekinesis team has already deployed using the skill library.

Relay Soldering

Laser Engraving

Assembly

Carton Palletizing

Quality Control

Automated Basil Harvesting

Quality Control (Panda)

Gear Assembly

Machine Tending

Simulation Applications ​

Develop and simulate digital twin workflows to validate, stress-test, and optimize Skill Groups. Deploy the same Skill Groups to real-world robots using a simulation-to-real transfer pipeline.

CNC Machine Tending

Surface Polishing

Pick and Place

Metal Palletizing

Robotic Welding

Palletizing

Physical AI Agents ​

Recent advances in LLMs and VLMs, including systems such as LLama 4, Mistral, Qwen, Gemini Robotics, RT-2, π₀, world models, and Dream-based VLAs have shown the potential of learned models to perform semantic reasoning, task decomposition, and high-level planning from vision and language inputs.

In the Telekinesis library, a Physical AI Agent, typically a Vision Language Model (VLM) or Large Language Model (LLM), autonomously interprets natural language instructions and generates high-level Skill plans. In autonomous Physical AI systems, Agents continuously produce and execute Skill plans, allowing the system to operate with minimal human intervention.

To learn more about building the Telekinesis Physical AI Agents, explore Cortex.

Skills ​

A Skill is a reusable operation for robotics, computer vision, and Physical AI. Skills span 2D/3D perception (6D pose estimation, 2D/3D detection, segmentation, and image processing), motion planning (RRT*, motion generators, trajectory optimization), and motion control (model predictive control, reinforcement learning policies). Skills can be chained into pipelines to build real-world robotics applications.

Below are examples of what a Skill looks like:

Example 1: segment_image_using_sam. This skill performs segmentation on an image using SAM model.

python
# Example 1
from telekinesis import cornea                                # Import Cornea - Image segmentation module

# Executing a 2D image segmentation Skill                     
result = cornea.segment_image_using_sam(                      # Executing Skill - `segment_image_using_sam`
    image=image,
    bboxes=[[400, 150, 1200, 450]]
)
# Access results
annotations = result.to_list()

Example 2: detect_objects_using_rfdetr. This skill performs object detection on image using RFDETR.

python
# Example 2
from telekinesis import retina                                   # Import Retina - Object detection module

# Executing a 2D object detection Skill
annotations, categories = retina.detect_objects_using_rfdetr(    # Executing Skill - `detect_objects_using_rfdetr`
    image=image,
    score_threshold=0.5,
)
# Access results
annotations = annotations.to_list()
categories = categories.to_list()

Skills are organized in Skill Groups:

Cornea - Image segmentation skills
python
from telekinesis import cornea
  • Color-based segmentation: RGB, HSV, LAB, YCrCb
  • Region-based segmentation: Focus region, Watershed, Flood fill
  • Deep learning segmentation: BiRefNet (foreground), SAM
  • Graph-based segmentation: GrabCut
  • Superpixel segmentation: Felzenszwalb, SLIC
  • Filtering: Filter by area, color, mask
  • Thresholding: Global threshold, Otsu, Local, Yen, Adaptive, Laplacian-based

See all the Cornea Skills.

Retina - Object detection skills
python
from telekinesis import retina
  • Classical shape detection - Hough Transform, Contours
  • 2D Object detection - YOLOX, RF-DETR
  • Open-Vocabulary detection - Qwen-VL, Grounding DINO

See all the Retina Skills.

Pupil - Image processing skills
python
from telekinesis import pupil
  • Morphology: erode, dilate, open/close, gradient, top-hat
  • Structure: Frangi, Hessian, Sato, Meijering
  • Edges: Sobel, Scharr, Laplacian, Gabor
  • Denoising: Gaussian, median, bilateral, box filters
  • Enhancement: CLAHE, gamma correction, white balance
  • Transforms: pyramids, mask thinning

See all the Pupil Skills.

Vitreous - 3D point cloud processing skills
python
from telekinesis import vitreous
  • Point cloud: centroids, normals, bounding boxes, principal axes
  • Filtering: masks, outliers, downsampling, plane & cylinder removal
  • Segmentation: DBSCAN, density, color, plane-based clustering
  • Transforms: rigid transforms, scaling, projection
  • Registration: ICP (P2P, P2Plane), global registration, cuboid sampling
  • Meshes: shapes, mesh to point cloud, convex hull, Poisson reconstruction

See all the Vitreous Skills.

Illusion - Synthetic data generation skills
python
from telekinesis import illusion
  • Synthetic image data generation for AI model training
Iris - AI Model training and deployment skills
python
from telekinesis import iris
  • AI model training pipelines
  • Fine-tuning and evaluation of foundation models
Neuroplan - Robotics skills
python
from telekinesis import neuroplan
  • Kinematics
  • Motion planning
  • Control
  • Robot database
Cortex - Physical AI Agents
python
from telekinesis import cortex
  • Action & Skill graphs
  • Physical AI Agents for skill graph generation and execution
RLBotics - Reinforcement Learning Skills
python
from telekinesis import rlbotics
  • Train RL policies in simulation (locomotion, manipulation, control)
  • Simulate and validate policies before deployment
  • Sim-to-sim: Deploy the same policy across different simulators
  • Sim-to-real: Transfer policies from simulation to real robots

See all the RLBotics Skills.

Robotics Skills ​

Using the skill group called neuroplan, prototype on any robot(industrial, mobile, or humanoid robot), perform any task on the same platform, and deploy the same Skill Groups anywhere - any robot, any task, on one Physical AI platform.

Robot Arm 1

Robot Arm 2

Robot Arm 3

Mobile Robot 1

Mobile Robot 2

Mobile Robot 3

Humanoid 1

Humanoid 2

Humanoid 3

One of the biggest pains of robotics is that each robot provider has their own interface to control their robots.

Use our library to run the same Skill Groups to interact with the leading industrial and mobile robots: Universal Robots (real & simulation), KUKA (real & simulation), ABB (real & simulation), Franka Emika (real & simulation), Boston Dynamics (simulation), Anybotics (simulation), Unitree (simulation).

Import robotic skills like shown below:

python
from telekinesis import neuroplan # robotics skills

Explore the full Neuroplan robotics stack in Neuroplan Overview.

Computer Vision Skills ​

Use the production-grade computer vision Skill Groups for obstacle detection, ground navigation, pose estimation, point cloud processing, bin picking, conveyor systems and AI model training.

3D Object Detection & 6D Pose Estimation

Mesh Processing

3D Point Cloud Registration

Forklift and warehouse obstacle detection using image segmentation for robotics and Physical AI

Obstacle Detection

Human detection and person identification using YOLOX object detection for robotics safety

Human Identification

Ground and floor segmentation for robot navigation and autonomous path planning

Ground Navigation

Bin picking in robot vision with segmented parts for industrial automation and grasping

Bin Picking

Depalletizing boxes with computer vision segmentation for warehouse and logistics robotics

Depallitizing Boxes

PCB circuit board segmentation for electronics manufacturing and automated inspection

PCB Segmentation

Object tracking on conveyor belt using open-vocabulary Grounding DINO detection

Conveyor Tracking

Object counting with AI vision for inventory management and quality control

Object Counting

Parts inspection using Hough circle detection for manufacturing quality control

Parts Inspection

Skills can be imported like shown below:

python
from telekinesis import cornea    # image segmentation skills
from telekinesis import retina    # object detection skills
from telekinesis import pupil     # image processing skills
from telekinesis import vitreous  # point cloud processing skills
from telekinesis import iris      # AI model training skills

Furthermore, we offer medulla which is a unified interface to cameras such as: Zivid, Mechmind, Microsoft Kinect, and others.

python
from telekinesis import medulla   # sensor interface skills

Learn more in the Medulla Overview.

Synthetic Dataset Generation Skills ​

Generate photo-realistic synthetic datasets to train and validate computer vision models. Replace months of manual data collection with our Illusion module.

Synthetic training dataset for computer vision and robotics - photorealistic industrial scene 1

Synthetic Data 1

Synthetic training dataset for computer vision and robotics - photorealistic industrial scene 2

Synthetic Data 2

Synthetic training dataset for computer vision and robotics - photorealistic industrial scene 3

Synthetic Data 3

Synthetic training dataset for computer vision and robotics - photorealistic industrial scene 4

Synthetic Data 4

Synthetic training dataset for computer vision and robotics - photorealistic industrial scene 5

Synthetic Data 5

Synthetic training dataset for computer vision and robotics - photorealistic industrial scene 6

Synthetic Data 6

Synthetic training dataset for computer vision and robotics - photorealistic industrial scene 7

Synthetic Data 7

Synthetic training dataset for computer vision and robotics - photorealistic industrial scene 8

Synthetic Data 8

Synthetic training dataset for computer vision and robotics - photorealistic industrial scene 9

Synthetic Data 9

Easily call the skill like shown below:

python
from telekinesis import illusion   # synthetic data generation skills

Find out more about the synthetic dataset generation skills in Illusion Overview.

Reinforcement Learning Skills ​

Train and deploy reinforcement learning (RL) policies for robotics with the Telekinesis Agentic Skill Library. From simulation to real hardware, RLBotics brings learned behaviors—locomotion, manipulation, and control—into your Physical AI pipelines.

Training Jump

Sim-to-Sim Deployment

Training Walk

Access the skills like shown below:

python
from telekinesis import rlbotics   # reinforcement learning skills

Learn about all the ways to train your reinforcement learning policies in RLBotics Overview.

BabyROS Middleware ​

BabyROS is a lightweight robotics framework built on Zenoh that implements the familiar Publisher/Subscriber and Client/Server architecture from ROS, without heavy system dependencies or complex middleware setup.

It is an open-source (Apache 2.0 License) middleware for communication between sensors, actuators, AI modules, and control algorithms, enabling reliable, low-latency messaging both on a single device and across distributed systems, from microcontrollers to cloud-based AI services.

BabyROS is designed for physical AI and robotics applications, including:

  • Robot perception and control pipelines: orchestrate high-frequency sensor data (cameras, LIDAR, IMUs) with actuator commands in real time.
  • Edge and distributed robotics: coordinate multiple devices, microcontrollers, and cloud modules with low-latency, high-throughput communication.
  • Rapid prototyping: iterate quickly on algorithms and behaviors without defining custom message types or managing workspace overlays.
  • Resource-constrained platforms: deploy on single-board computers, embedded systems, or microcontrollers where ROS installation is impractical.
  • High-throughput, shared-memory communication: efficiently transfer large or frequent datasets within the same machine.

Who Is Telekinesis For? ​

  • Robotics startups & product teams: Ship faster by using production-grade perception, motion planning, control, and reinforcement learning — without stitching together fragmented stacks.

  • Industrial automation engineers & integrators: Deploy sim-to-real robotic systems with a unified Python SDK that works across industrial, mobile, and humanoid robots.

  • AI & computer vision researchers: Prototype, validate, and deploy Physical AI pipelines — from 6D pose estimation to reinforcement learning policies — in one consistent framework.

Join our Discord community to exchange ideas, contribute Skills, and accelerate the development of real-world robotics systems.

Developer Quickstart ​

You can easily integrate Telekinesis Agentic Skill Library into your own application. Setup the library in just 4 steps and start building!

Step 1: Generate an API Key ​

Since all the skills are hosted on the cloud, to access them securely, a free API key is needed. Create a Telekinesis account and generate an API key for free: Create a Telekinesis account!

Store the key in a safe location, such as your shell configuration file (e.g. .zshrc, .bashrc) or another secure location on your computer.

Step 2: Configure the API Key ​

Export the API key as an environment variable. Open a terminal window and run below command as per your OS system.

Replace <your_api_key> with the one generated in Step 1.

bash
export TELEKINESIS_API_KEY="<your_api_key>"
powershell
setx TELEKINESIS_API_KEY "<your_api_key>"

WARNING

For Windows, after running setx, restart the terminal for the changes to take effect.

The Telekinesis SDK uses this API key to authenticate requests and automatically reads it from your system environment.

Step 3: Pip Install the Telekinesis Agentic Skill Library ​

We currently support Python versions - 3.11, 3.12. Ensure your environment is in the specified Python version.

Install the core SDK using pip:

bash
pip install telekinesis-ai

Step 4: Run Your First Example ​

  1. Git clone our telekinesis-examples repository from Github with:
bash
git clone --depth 1 --recurse-submodules --shallow-submodules https://github.com/telekinesis-ai/telekinesis-examples.git

INFO

This also downloads the telekinesis-data repository, which contains sample data used by the examples. You can replace this with your own data when using Telekinesis in your own projects. Download time may vary depending on your internet connection.

  1. Change directory into telekinesis-examples:
bash
cd telekinesis-examples
  1. Install example-only dependencies:
bash
pip install numpy scipy opencv-python rerun-sdk==0.27.3 loguru pycocotools
  1. Run the segment_image_using_sam example:
bash
python examples/cornea_examples.py --example segment_image_using_sam

If the example runs successfully, a Rerun visualization window will open showing the result.

INFO

Rerun is a visualization tool used to display 3D data and processing results.

Input-Output Visualization

SAM image segmentation input-output demo - Telekinesis Cornea quickstart example

Join our Discord Community to Add your Own Skills ​

The Telekinesis Agentic Skill Library is the beginning of a vibrant ecosystem. Whether you are a researcher, a hobbyist, or an industrial engineer, your work belongs here. Release your Skill, let others improve it, and see it deployed in real-world systems.

Join our Discord community to be part of the Physical AI revolution!

What Next? ​

Explore all the features of Telekinesis Agentic Skill Library with Cornea!