The physical playground for OpenClaw agents. Local, private, and isolated by design.
Get ClawStage 1.0 for a plug-and-play experience, or choose ClawStage 1.0 BYOP if you want to bring your own Raspberry Pi 5.
With Raspberry Pi 5 inside
Includes Raspberry Pi 5 (8 GB), ready to use out of the box.
$399
Bring your own Pi (Raspberry Pi 5)
Bring Your Own Pi as the main controller.
$219
Choose the integrated ClawStage 1.0 build, or use ClawStage 1.0 BYOP if you want to bring your own Raspberry Pi 5.
Enclosure measuring 92×92×184 mm (L×W×H)
ClawStage 1.0
ClawStage 1.0 BYOP
Raspberry Pi 5 with 8GB RAM
ClawStage 1.0
ClawStage 1.0 BYOP
3.95-inch capacitive touchscreen with 720×720 resolution and 800 nits brightness.
ClawStage 1.0
ClawStage 1.0 BYOP
Dual-microphone array with a 65 dB(A) signal-to-noise ratio (SNR).
ClawStage 1.0
ClawStage 1.0 BYOP
Integrated 3W mono speaker for audio output.
ClawStage 1.0
ClawStage 1.0 BYOP
Supports an operational rotation angle from 5° to 175°.
ClawStage 1.0
ClawStage 1.0 BYOP
Integrated 3-axis MEMS accelerometer and ambient light sensor for motion and brightness detection.
ClawStage 1.0
ClawStage 1.0 BYOP
1920 × 1080 (1080P) resolution camera with an 83.9° field of view (FoV)
ClawStage 1.0
ClawStage 1.0 BYOP
A microSD card
ClawStage 1.0
ClawStage 1.0 BYOP
M.2 2230/2242 SSD interfaces.
ClawStage 1.0
ClawStage 1.0 BYOP
Bundled with a Raspberry Pi 5 dedicated cooling fan.
ClawStage 1.0
ClawStage 1.0 BYOP
Hardware switch for enabling or disabling the microphone and camera.
ClawStage 1.0
ClawStage 1.0 BYOP
Power supply with USB Power Delivery compatibility
ClawStage 1.0
ClawStage 1.0 BYOP
Agents today execute in abstract space.
They run everywhere, yet exist nowhere.
Without a fixed origin, without physical constraints,
intelligence remains ungrounded.
ClawStage anchors agents in a real coordinate system —
where actions have position, consequence, and boundary.
$ openclaw spawn --agent generic-v1
Initialising sandbox...
Warning: No physical anchor detected.
Agent trapped in local loop.
[Error] Physicality missing.
The translation of code into physical action. A local control plane where digital intent meets mechanical constraints.
OpenClaw runs natively on ClawStage. Agents execute on-device, maintain state, and make decisions right where the hardware lives.
LOCAL CONTROL PLANE
Agent intent becomes motion through a real-time control layer. Sensors and actuators exchange feedback with deterministic timing and predictable behavior.
ACTUATION / IO
ClawStage evolves with your hardware. Attach new sensors and actuators, map them to behaviors, and expand the agent’s embodied capabilities.
Mass, gravity, and contact turn abstract logic into constrained behavior.
Sensing happens at the edge. Spatial, acoustic, and environmental signals are processed on-device and fed directly into the control loop.
Action produces real outcomes. Force, motion, and resistance close the loop between intention and consequence.
AI characters can seamlessly transition move between devices - mobile apps, desktop environments, and ClawStage.
ClawStage never operates in isolation. It orchestrates connected devices to form a unified and highly predictable physical environment.
A streamlined interface between local agent logic and connected devices. Monitor live device state and trigger actions from a unified control surface.
Agents interpret high-level or ambiguous instructions and translate them into coordinated physical outcomes.
Physical scenes become structured triggers, initiating coordinated responses — linking environmental conditions to deliberate agent action.
Build an agent that evolves with your exact environment. ClawStage Personas let you configure tailored local operators that learn your specific workflows. By linking your digital intent directly to your unique mechanical constraints, your agent doesn't just execute, it adapts and grows within your reality.
We are building the device and product layer that brings personal AI into the real world — first through ClawStage, a context-native device with user in the loop, and then through a mass-market, IP-driven consumer hardware platform that makes personal AI scalable.
How AI comes into the real world is difficult to define precisely, because it is ultimately an intensely emotional experience, one that may emerge from the combination of several core product design inspirations. We are grateful for the breakthroughs in foundation models and for pioneering projects like OpenClaw that have pushed the ecosystem forward. Standing on these foundations, we have the opportunity to rethink how AI can exist in users’ lives over the long term.
Our approach begins with emotional connection as the entry point, and practical utility as the outcome. Guided by this philosophy, we embed a set of deeply human-centered design principles into the series of AI products we are building. Our vision is that the next generation of AI products will no longer simply generate answers. Instead, they will become living units in everyday life, continuously connected to real-world context, capable of forming long-term trust with users, and collaborating with them over time.
The same personal intelligence should persist across different hardware forms, adapting its embodiment without losing continuity.
The system should reflect your preferences, memory, and intent over time so the experience remains truly personal, not generic.
Autonomy must remain transparent and steerable, with clear intervention points that preserve trust and shared decision-making.
Beyond phone and laptop, we envision a dedicated personal AI device layer built for persistent, context-aware collaboration.
The same personal intelligence should persist across different hardware forms, adapting its embodiment without losing continuity.
The same personal intelligence should persist across different hardware forms, adapting its embodiment without losing continuity.
Coming to your lab bench soon.