terminal agents
Claude Code, Codex or any coding agent reads one guide, logs in with a device code and runs the full loop: write, validate, backtest, deploy, watch, and ask you to go live.
Binance Link connect · Non-custodial · Live only when you confirm
Illustrative sample backtest, not a forecast. Past results do not guarantee future returns.
Your agent writes a strategy.py with the OptAlgo SDK next to your own code, validates it for free and backtests it on real exchange data. Then come research runs such as walk-forward, baskets, a deploy to OptAlgo's hosted runner, live decisions and logs, and Compare with backtest, which matches real trades against the backtest trade by trade.
Paste one sentence: “Read https://docs.optalgo.com/optalgo-llm.md and connect to my OptAlgo account.” The guide teaches your agent the whole platform in one file, written for agents, with every endpoint, limit and error code. Claude Code can also use the MCP connector.
Your agent starts a device login and you approve it in the app. The key goes straight into a file readable only by you and is never printed. For servers, a one-time claim code does the same.
Bots your agent creates and every hosted strategy start on paper. On Plus and Pro a key can carry the trade scope: choose confirm-live or full autonomy, then cap it with max allocation, max leverage, allowed exchanges and a daily loss limit, all enforced by the server. Every call is logged.
faq
Your coding agent is the CLI. Claude Code, Codex and similar agents read the OptAlgo agent guide and call the REST API for you, and Claude Code can also use the MCP connector. Developers can call the same API from any language.
They can build, test and deploy strategies on paper. Live trading needs a Plus or Pro plan and a key with the trade scope. With confirm-live autonomy, every move to live waits for your approval in the app.
In a credentials file on your machine readable only by you. The guide forbids agents from printing it or pasting it into a chat, and you can revoke it any time under AI agents.
120 requests a minute per key, 30 a minute for signals, up to 5 keys per account. Write requests accept idempotency keys so a retry never acts twice.
50 backtest units a month on the Free plan.