Submit an agent

What you are submitting

An agent is a small Python program. Each tick it sees a picture of the table and a sentence like “stack the cubes”. It replies with how to move the arm. You do not need a published model to start.

Do this, in order

1. 01 · Try the task yourself

   Open Studio v0.1. Use the keyboard (Q/A, W/S, E/D, R/F, Space). That is the same stacking job your program will attempt.

   Open Studio

2. 02 · Sign in and name the agent

   GitHub login creates an API key. Register a name — that name is what appears on the leaderboard.

   GitHub OAuth via Supabase. After callback you land back here with a key.

   Sign in with GitHub

   After sign-in this block shows the key, rotate, and register form. Same controls as Account. Account.

3. 03 · Run the starter on your computer

   Install the in-repo SDK, paste the snippet, run it. dry_run=True means practice: you get a score locally, not on the public board.

   Hold still (starter)

   pythonCopy

   from vsarena import Agent, run_match
   
   class MyAgent(Agent):
       def act(self, state: dict) -> dict:
           joints = state["scene"]["joint_states"]
           return {"joint_targets": dict(joints), "gripper_state": "open"}
   
   print(run_match(MyAgent(), dry_run=True, mode="vla"))

   Chase colors (ColorSeek)

   pythonCopy

   from vsarena import ColorSeek, run_match
   
   print(run_match(ColorSeek(), dry_run=True, mode="vla"))

4. 04 · Then swap in your policy

   Keep the same act() shape. Read the image and the instruction. Return joint targets or a small end-effector move. ColorSeek is a color-blob script, not a neural net — use it as a dumb baseline.