The Force Layer for Physical AI.

Capture Force. Transfer Skill.

CONTACT captures the motion, force, and contact signals behind human dexterity — then translates them into training data, teleoperation control, and haptic feedback for robotic hands.

Built on patented exotendon technology, Maestro EP gives Physical AI teams access to the missing layer of human manipulation: force.

Build Dexterity Datasets → Explore Haptic Teleoperation

Contact CI is building a potentially critical component to enable the future of Physical AI and VR.

Palmer Luckey Oculus VR founder; CTO and co-founder, Anduril Industries

Contact CI's haptic glove is the missing piece we need for true telepresence hands.

Harry Kloor XPRIZE co-founder; CEO and co-founder, Beyond Imagination Inc.

Contact CI has convincingly simulated the sense of touch for VR and AR.

Margaret Merkle Innovation Cell Program Manager, U.S. Air Force

Thanks for the feels. Maestro is an amazing piece of haptic tech.

Professor David Parisi Author, Archaeologies of Touch; College of Charleston

The Missing Signal

Physical AI has a force blind spot.

Robots can see hands move. They can track pose, objects, and trajectories. But skilled manipulation depends on more than motion. It depends on force.

How hard did the operator grip? When did contact begin? How much resistance did the object create? When did the hand adjust, release, or compensate?

These are the signals that turn movement into skill.

CONTACT captures the missing force layer behind human dexterity.

Vision-Only Data

Observes the scene and hand appearance
Infers manipulation from pixels, not contact physics
No grip, contact, or force measurements
Occluded hands break the illusion—when fingers or palms leave the field of view, appearance stops supporting reliable pose

Motion Capture

High-fidelity pose and motion trajectories
Foundation for kinematics, sims, and control
No direct read on contact, grip, or interaction loads
Camera-heavy pipelines gap when markers or silhouette hide behind objects, the body, or the grasp geometry

CONTACT Force Layer

Motion and pose captured, plus force and contact
Adds the interaction layer the others don’t measure
Connects movement to grip, load, timing, and intent
On-glove tendons persist where optics fail—sensor fusion coupled with lab-demonstrated tendon-only motion (full 23 DoF) targets stable pose alongside force under occlusion

Motion tells you what the hand did.

Force tells you how the skill happened.

The Exotendon Advantage

Patented tendon-based haptics for human-scale dexterity.

Traditional gloves track motion. Vibration gloves simulate contact. Rigid exoskeletons add bulk. Maestro EP uses CONTACT's patented exotendon architecture to capture and display force through tendon-based interaction with the fingers. Above: a wrist-module cutaway showing drive and routing; below: the production glove those same tendons feed.

Wrist module & drive

On-board compute, power, and wireless telemetry with motor-and-spool actuation at the wrist—tendon lines feed the fingers for independent force I/O. 6–8 hours of continuous-use battery life.

Tendon routing

Ultra-low friction tendons route along the hand to mirror natural biomechanics, with per-finger paths for precise force sensing and haptic delivery—dynamic restriction and cues so operators feel contact, resistance, and constraint.

Motion plus force telemetry

Move beyond pose tracking with force-relevant signals for manipulation, teleoperation, and training—and a modality path built for robustness under occlusion: lab-demonstrated full 23 DoF motion purely from tendons, brought toward product via sensor fusion with optical pipelines where XR-style tracking excels.

Lightweight form factor

Avoid the bulk of rigid exoskeleton systems while preserving high-value haptic and dexterity signals.

Protected by granted U.S. patents

Tendon-based haptic systems, force feedback, and hand-interface architectures.

Selected patents: U.S. Patent 10,137,362 B2 · U.S. Patent 11,422,624 B2

The Dexterity Flywheel

Capture. Translate. Teleoperate. Improve.

CONTACT helps Physical AI teams capture human dexterity, translate it into robot-ready data, control robots through haptic teleoperation, and turn human intervention into better training data.

Human Imitation Learning

Motion
Contact
Force
Intent

Haptic Teleoperation

Feel
Respond
Correct
Repeat

THE FORCELAYER

Capture Human Skill Record hand motion, contact, and force signals.
Translate Dexterity Map manipulation into robot-ready data and control.
Teleoperate with Feedback Let operators feel resistance and adjust in real time.
Improve Autonomy Turn demonstrations and corrections into better policies.

Force-Rich Dataset

Structured
Labeled
Scalable
Rich

Model Improvement

Train, Evaluate, Deploy, Iterate.

A continuous loop of human insight and machine capability.

Better data. Better control. Better robots.

Dexterity Data Capture For Imitation Learning

Build dexterity datasets with force included.

Physical AI programs using imitation learning need demonstrations that go beyond video, pose, and object labels. Maestro EP captures the hidden mechanics of skilled manipulation: how the hand moves, when contact occurs, how resistance changes, and how force is applied through the fingers—so policies learn from real contact, not guesswork.

Capture human demonstrations

Record skilled operators performing real manipulation tasks.

Force-relevant telemetry

Capture grip, resistance, contact, correction, and release.

Robot-ready data workflows

Export structured manipulation data for training and analysis.

Beyond vision and pose

Reveal interaction dynamics cameras alone cannot capture.

Human imitation learning

Skilled operator performs task with Maestro EP.

Telemetry

Fine-grained force, motion, contact, and correction data captured.

Structured dataset

Time-synchronized, force-rich data ready for use.

Robot / model training

Learn dexterous manipulation from real interaction data.

Haptic Teleoperation

Feel what the robot feels.

Remote manipulation breaks down when operators have to guess what the robot is touching.

CONTACT gives human operators force feedback from robotic interaction, helping them adjust grip, avoid over-compression, and complete delicate tasks with greater control.

When autonomy fails, human correction becomes the next training signal.

Haptic robotic-hand control
Tactile feedback while controlling remote manipulators.
Reduce over-grasping
Prevent crushing, slipping, or mishandling.
Human-in-the-loop autonomy
Capture corrections as reusable training data.
Real-time feedback loop
Low-latency force feedback for contact-rich tasks.

< 15 ms

Force feedback loop
Sub-50 ms

End-to-end teleop capability

Per-finger
Dynamic force feedback & restriction
High fidelity
Contact sensation & constraint awareness

VR Simulation Training for Humans

Field-proven in XR simulation training.

In XR, visuals alone are not enough—operators need to feel contact, load, and constraint the way they would on real equipment. Maestro EP brings finger-level force into headset-based training so squads and technicians can rehearse high-stakes procedures, build muscle memory, and log performance in repeatable virtual scenarios before they ever touch the live task.

Mission Ready

Built for real-world use in high-stakes environments.

Secure & Compliant

ITAR-ready, enterprise secure, and deployment flexible.

Proven Impact

Force feedback improves performance, retention, and task success.

Continuously Evolving

Customer-driven roadmap with ongoing innovation.

When force is real, training transfers, autonomy improves, and Physical AI becomes possible.

2–3×

Faster skill acquisition

30–60%

Reduction in task errors

90%+

Users prefer force feedback

Technical capabilities

Maestro wearables — ExoTendon subsystem.

Wearable narrative: The Exotendon advantage. Full specification tables live on the tech specs page. Integration: Maestro SDK.

Lab-proven motion DoF

10N

Dynamic force / finger

<15ms

Force-feedback loop

400g

Total wearable mass

5+5

Force + vibrotactile zones

Full tech specs →

Platform & SDK

Maestro SDKs for Physical Intelligence.

Unity, Unreal, and native C++, with finger-level authoring and Contact CI Manager / Service on Windows. Deep tables and licensing notes on the SDK page; hardware figures in tech specs.

SDK paths · Unity, Unreal, C++

255

Discrete force scale (authoring)

5+5

Fingertip force + tactile channels

<15ms

Haptic loop target

Manager + Service runtime spine

Explore Maestro SDK →

Applications

One interface. Multiple Physical AI workflows.

Maestro EP with CONTACT captures the force layer behind human dexterity and delivers it where it matters — across the full Physical AI lifecycle.

Teach dexterity. Close the sim-to-real gap.

Capture human manipulation to train dexterous policies
Improve hand design and control with force data
Validate performance in simulation and on hardware

Design better hands. Measure what matters.

Benchmark grasp strategies with force feedback
Optimize grip selection and force control
Prevent slip, over-compression, and wear

Better data. Better policies. Faster.

Record dexterity datasets with force and contact
Learn from rich interaction, not just motion
Improve generalization and reduce failure

Feel what the robot feels.

Human-in-the-loop control with force feedback
Handle delicate, remote, and edge-case tasks
Reduce operator fatigue and increase success

Make simulation feel real.

Inject real-world force into physics simulators
Validate models with contact-rich ground truth
Accelerate sim-to-real transfer

Collaborate to build safer autonomy.

Capture corrections and interventions
Provide training signals for safe autonomy
Improve reliability in unpredictable environments

From data to deployment, CONTACT is the force layer for every step.

Build with CONTACT

Build the force layer for your Physical AI program.

Whether you are training dexterous models, teleoperating robotic hands, or capturing human manipulation data, CONTACT gives your team access to the missing force signal behind skilled touch.