> ## Documentation Index
> Fetch the complete documentation index at: https://servo.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Servo

> Run hosted robot policy models on a Bimanual YAM.

Servo runs robot policy models on managed compute and returns action chunks to the computer wired
to your Bimanual YAM.

You provide a zero-argument function that opens the calibrated YAM runtime from your robot
application. Servo reads cameras and joint state through that runtime and sends validated joint
targets back through it. Servo handles model compatibility, deployment, image encoding, hosted
inference, and action timing.

The workflow has four steps:

1. Register each physical YAM under a stable name.
2. List the models compatible with that robot.
3. Deploy the model through Servo.
4. Attach the existing runtime, check the full path without sending actions, and start a timed run.

Start with [Run pi0.5](/quickstart). Use [Robots](/guides/robots) for the camera, joint, actuator,
and calibration contract. Use [Robot fleets](/guides/fleet-deployment) when several robots share
hosted capacity.