AI Trading Agent

Rent a hosted agent runtime and drive your own O2 account in plain language.

What this is

A container running open-source opencode with the O2 trading SKILL preloaded. You tell it what you want; it calls the O2 API on your behalf — using an API key you created, with funds and positions staying in your own account.

opencode

DEEPSEEK-V4PRO

O2 trading SKILL preloaded

Three steps

Create an API key

Create a key with trade permission. The secret is shown once — save it immediately.

Create one

Start the agent

Inject your key and the model config into the container. On startup it checks all three links and names the failing one.

Trade in plain language

"Show me the BTC book" / "Put a 3% stop on my long" — the agent reads account state before acting.

What the agent can and cannot do

This list reflects real backend permission gates, not settings — the server refuses regardless of what the agent attempts.

Can

Read markets, order book, positions, balance and account equity
Open positions (limit / market)
Close positions and cancel orders
Set and cancel take-profit / stop-loss
Fetch historical candles

Cannot

✕Withdraw — API keys are always refused; you must sign in yourself
✕Create or revoke API keys — otherwise a read-only key could escalate itself
✕Touch funds outside the main account

Rent a runtime

Coming soon

Container and model quota, rented together. Pricing and terms are not set yet; they will appear here once they are — no estimated figures until then. You can prepare your API key now.

Configuration

Once you have a key, start the agent container with this command.

docker run -d --name o2-agent \
  -e O2_API_KEY_ID=<your key id> \
  -e O2_API_SECRET=<the secret you saved> \
  -e O2_RELAY_BASE_URL=<issued to you> \
  -e O2_RELAY_API_KEY=<issued to you> \
  -p 127.0.0.1:4096:4096 o2-agent-runtime:1

The secret is returned only at creation and stored as a hash — it cannot be filled in for you. The two relay values are issued with your container.