vla-arms-agent

See, decide, act.

Connect cameras and a robot arm. Give tasks in natural language.

Follower disconnected
01

Connect the arm and bind external and wrist cameras.

Follower connection

Read the current pose after connecting.

Execution state
Disconnected
Actual gripper opening
—
Gripper power feedback
—
Joint mapping and offsets

Direction and offsets calibrate joint angles. Changes update the pose and preview.

JointDirectionOffset / °
J1
J2
J3
J4
J5
J6

External camera

Off
Select a device and start preview

Wrist camera

Off
Select a device and start preview

If video stutters, close the camera you are not using.

02

AI and tasks

Keep context and send reminders while a task runs.

Waiting for a task
Model and motion settings

The model needs image understanding and tool calling. Changes apply to the next task. Gripper power 0 uses its default.

Describe a task

Example: inspect both views, move the purple sponge to the center mark, then return to the work pose.

With the external preview active, its current image is attached to the task.

Pose and manual control

Actual feedbackTarget preview
XYZConnect the follower to see its actual pose
J1—
J2—
J3—
J4—
J5—
J6—
Actual TCP / mm—
Joint targets and gripper

The work pose closes the gripper. Opening references: about 71 mm at 30° and 129 mm at 60° (measured at different depths).

TCP position IK

Enter a target position, preview it, then execute. If IK fails, adjust the joint configuration first.

Actual servo feedback
IDRaw angleVoltageCurrentPowerStatus

Continue this conversation?

This uses the existing chat history. Start a new conversation for a new task.