New Product Launch: Agility A2 Flex Brings Mobile Manipulation and Embodied AI Research Together

As robotics moves beyond fixed workstations, developers are increasingly looking for systems that can navigate different environments, interact with objects, and perform coordinated manipulation tasks. However, turning these capabilities into a practical robotic solution remains a challenge, especially when mobility, perception, and arm control must work together.

Foxtech introduces the new Agility A2 Flex, a 27-DOF open-source wheeled dual-arm robot designed for mobile manipulation, whole-body VR teleoperation, and embodied AI research.

Combining a multidirectional mobile base, adjustable lifting mechanism, dual robotic arms, and an integrated control architecture, Agility A2 Flex gives researchers and developers a unified platform for exploring more advanced robotic behaviors.

Available in Agility A2 Flex and Agility A2 Flex Ultra configurations, the platform supports different development needs, from human-guided teleoperation and motion control to VLA model training and validation.

Why Do Robotics Developers Need More Than a Fixed Robotic Arm?

Traditional robotic arms are effective for tasks performed within a defined workspace. However, real-world environments are rarely limited to one fixed position.

A robot working in a laboratory, warehouse, or experimental manipulation environment may need to approach objects from different directions, adjust its working height, and coordinate movements between both arms. These requirements introduce challenges that cannot be solved by robotic arm control alone.

Agility A2 Flex addresses these limitations by combining mobility and manipulation within a single robotic system.

Its 27-DOF architecture integrates dual arms, grippers, a movable head, vertical lift, and mobile base. This enables developers to study how different robotic components can work together to complete more complex operations.

Rather than focusing only on individual joint movements, researchers can develop and evaluate whole-body robotic behaviors.

How Can Agility A2 Flex Adapt to Different Working Environments?

One of the key challenges in mobile manipulation is reaching objects located at different positions and heights.

Agility A2 Flex features a 4W4S (Four-Wheel Drive, Four-Wheel Steering) mobile base that supports multidirectional movement. This allows the robot to reposition itself without being restricted to conventional forward-and-backward driving patterns.

The integrated 700 mm vertical lifting mechanism further expands its working flexibility by allowing the robotic arms to operate at different heights.

Combined with dual 7-joint robotic arms, each offering an approximate 714 mm operating range, the platform supports coordinated grasping, handling, and object transfer.

These capabilities are particularly useful for developing mobile manipulation workflows involving tables, shelves, containers, and other objects positioned throughout a workspace.

For robotics teams, this means more opportunities to test manipulation strategies under changing environmental conditions rather than relying exclusively on stationary demonstrations.

How Can Developers Test Robot Tasks Without Building Autonomous Algorithms First?

Developing autonomous robotic behavior requires extensive experimentation. Before training an AI model, researchers often need to understand whether a task can be physically performed and how different robotic components should coordinate.

Agility A2 Flex provides an accessible starting point through its integrated whole-body VR teleoperation system.

Using the included Meta Quest 3S VR headset, operators can control the robot's arms, grippers, head, lifting mechanism, and mobile base.

This allows researchers to demonstrate and evaluate manipulation tasks directly, without first developing a complete autonomous control policy.

For example, an operator can guide the robot through a coordinated object-handling sequence, observe how the robot responds, and identify areas where motion planning or control strategies may need improvement.

The same teleoperation capabilities also support human demonstration data collection, creating opportunities for subsequent robot-learning research.

By separating initial task validation from autonomous policy development, research teams can establish practical workflows before moving into more advanced AI experiments.

How Does an Open Development Platform Reduce Integration Challenges?

For universities, research laboratories, and robotics startups, software compatibility is often as important as hardware capability.

Developing a mobile robotic system from separate components can require substantial effort to connect controllers, sensors, motion planning frameworks, and communication interfaces.

Agility A2 Flex is built on Ubuntu 22.04 and ROS 2 Humble, providing an open development environment familiar to many robotics teams.

Both Agility A2 Flex and Flex Ultra support ROS 2 real-time joint control and MoveIt 2 arm motion planning.

These capabilities allow developers to work on robotic movement, manipulation planning, and customized control applications within an established robotics software ecosystem.

For research teams, an open development architecture also provides greater flexibility to experiment with different algorithms, modify control strategies, and expand the platform as project requirements evolve.

Can Agility A2 Flex Support the Transition from Teleoperation to AI-Driven Robotics?

Human-guided teleoperation is an important part of robotic development, but many research projects aim to move beyond remote control toward learning-based robotic behavior.

This transition requires a combination of demonstration data, environmental perception, computing resources, and AI model development tools.

Agility A2 Flex provides the foundation for teleoperation, motion control, and demonstration data acquisition. For teams working on more advanced embodied AI applications, Agility A2 Flex Ultra introduces additional capabilities.

The Ultra configuration includes dual Intel RealSense D405 wrist cameras, a Livox MID-360S LiDAR, an additional base-mounted RealSense D435i camera, and an NVIDIA RTX 3090 GPU.

It also supports Nav2 navigation and includes a VLA configuration using π0.5 as the default model.

With these additional components, researchers can explore how visual perception, robotic actions, and learning models interact during manipulation tasks.

The Ultra configuration supports VLA model training, inference, and workflow validation, making it suitable for teams investigating the transition from human demonstrations to AI-driven robotic control.

Which Configuration Is Right for Your Research Project?

Choosing a robotic platform depends on the project's development goals rather than simply selecting the highest hardware configuration.

For teams primarily interested in VR teleoperation, robotic motion planning, mobile manipulation experiments, and human demonstration collection, the standard Agility A2 Flex provides the essential hardware and software environment.

It includes an X86 controller, head-mounted Intel RealSense D435i camera, ROS 2 joint control, and MoveIt 2 motion planning.

For researchers who need richer perception data, mobile navigation, local GPU computing, and integrated VLA development capabilities, Agility A2 Flex Ultra offers a more comprehensive configuration.

The distinction is particularly important for customers planning long-term embodied AI projects. The standard Flex configuration does not include a VLA configuration, while Flex Ultra is specifically equipped for VLA training, inference, and validation.

By offering two configurations, the platform allows customers to select a development environment that better matches their current research objectives.

What Opportunities Does Agility A2 Flex Offer for Future Robotics Development?

Robotics research rarely remains limited to its initial requirements. A project that begins with remote manipulation may later expand into learning-based control, more complex object handling, or additional sensing capabilities.

Agility A2 Flex supports this progression through its expandable hardware architecture.

Developers can explore optional integration with the ExoArm-7 Pro wearable exoskeleton for alternative teleoperation methods. Optional dexterous hands can support research involving more complex grasping and manipulation.

The platform also supports custom end-effectors, vision devices, and OEM/ODM customization, providing additional flexibility for specialized development needs.

These options are particularly valuable for research institutions and engineering teams seeking a robotic platform that can evolve alongside their projects.

Optional ExoArm-7 Pro wearable exoskeleton and dexterous hands are sold separately.

Who Can Benefit from Agility A2 Flex?

Agility A2 Flex is designed for organizations developing and evaluating robotic intelligence.

Universities and educational institutions can use the platform to teach mobile robotics, manipulation control, and human-robot interaction through practical experiments.

Embodied AI laboratories can investigate coordinated dual-arm motion, human demonstration collection, and learning-based manipulation.

Robotics startups and industrial R&D teams can use the system for proof-of-concept validation, allowing them to explore robotic workflows before committing to more specialized development.

For teams requiring advanced perception and VLA capabilities, Agility A2 Flex Ultra provides additional resources for evaluating AI-driven robotic applications.

Conclusion: From Robotic Demonstration to Intelligent Mobile Manipulation

The development of capable mobile robots requires more than powerful hardware. It requires a coordinated system that allows researchers to control, observe, evaluate, and continuously improve robotic behavior.

The newly launched Agility A2 Flex brings together 27-DOF whole-body coordination, multidirectional mobility, dual-arm manipulation, VR teleoperation, and open-source development capabilities.

For developers beginning with teleoperation and robot control, Agility A2 Flex provides a flexible research foundation. For teams exploring advanced perception and VLA-based robot learning, Flex Ultra offers additional tools to support more demanding development workflows.

By connecting practical manipulation experiments with open robotics development, Agility A2 Flex helps research teams move from initial concepts toward more capable embodied AI systems.