AI Agent
The AI Agent understands your intent, takes action, and offers suggestions—handle strategies, backtests, market data, and bots through plain conversation.
AI Agent is Pluto’s in-system AI agent: it understands your needs, executes intelligently, and gives suggestions. It can orchestrate strategies, strategy optimization, backtests, quotes, and trading bots—while actual execution stays on the underlying trading system and risk rules. This page covers what it is, what it can do, typical scenarios, how to use it, and safety boundaries. For configuration, see AI Config.
Get started in one sentence
Say what you need in natural language; AI Agent infers intent, runs actions, and explains. For example: “Create a dual-MA strategy for me,” “Run a 3‑month backtest on BTC/USDT,” “What’s my P&L today?”—no page-hopping; professional trading workflows feel like chat.
What Is AI Agent
AI Agent (Pluto AI Agent) is the system’s unified assistant. It answers all kinds of quantitative-trading questions and, through conversation, helps you create strategies, run backtests, manage bots, and query markets. Whether or not you have quant experience, you can ask in plain language—AI Agent explains concepts, breaks down tasks, runs actions, and suggests next steps so you can get productive quickly and close the gap in trading experience and knowledge.
You don’t need to jump between pages: express ideas in natural language. AI Agent understands your intent, acts within the system’s safety boundaries, and returns explanations and suggestions.
Pluto uses a “AI Agent + trading system” two-layer architecture: AI Agent is the “brain” for understanding and orchestration; the underlying trading system performs real execution and validates against risk and permission rules. So AI is the decision and interaction layer; the system is the safe, controllable execution layer.
Leveling the experience gap
AI Agent can answer questions on strategies, backtests, risk, indicators, and more—even with zero quant background you can learn by asking and doing. From “What does Sharpe mean?” and “Why backtest before live?” to “How do I set take-profit/stop-loss?”, you can ask directly and build intuition fast.
The AI Agent acts as the "brain"—it understands natural language and dispatches the system tools below; every action is bounded by the underlying trading engine and risk controls.
What It Can Do
| Scenario | What it helps with |
|---|---|
| Strategies | Shape ideas, configure buy/sell conditions, explain logic, compare strategies; create from zero or customize from templates. |
| Backtest & optimization | Start backtests, read results and key metrics (Sharpe, drawdown, win rate), suggest parameter and config improvements; compare timeframes or pairs. |
| Live bots | Create and configure bots, explain runtime status, query positions and orders, debug “why no entry” or “why stopped out”; start/stop and tune parameters. |
| Quotes & positions | Real-time price and market context, trend and indicator interpretation, risk or opportunity hints; combine with positions for P&L and exposure. |
| System usage | Walk through Exchange Config, Notification Config, onboarding, and step-by-step guidance. |
All of the above runs inside the trading engine and risk rules: AI Agent cannot bypass position limits, stop-loss, or permissions; for live trading, funds, or important config changes it surfaces key parameters and asks you to confirm first.
Typical usage scenarios
Scenario examples
Below are common examples; real capabilities depend on your system. Multi-turn dialogue, context, and follow-ups are supported.
| Scenario | Example prompts |
|---|---|
| Quick strategy creation | “Build a MA5 / MA20 golden/death cross strategy,” “Create an RSI overbought/oversold strategy.” |
| Run a backtest | “Backtest Binance BTC/USDT spot for the last 3 months,” “Run backtest on 1H.” |
| Interpret results | “What does this Sharpe tell me?”, “Is 15% max drawdown high?”, “Why is win rate so low?” |
| Manage bots | “Start my XXX strategy bot,” “Today’s P&L on this bot?”, “Why hasn’t it opened a position?” |
| Market queries | “BTC price now?”, “How’s ETH been lately?”, “Show Binance BTCUSDT 1H candles.” |
| System setup | “How do I add an exchange?”, “Add Telegram notifications,” “First time here—how do I start fast?” |
Where to use it
| Entry | Description |
|---|---|
| Admin UI | Open AI Agent or the chat entry in the product (sidebar, top bar, or floating button on feature pages). Type natural language. Good for deep work and multi-turn flows with charts/tables when the UI returns them. |
| Group chat (DingTalk / Telegram) | With Notification Config configured and smart reply on, @ the bot or send messages in the group; AI Agent replies automatically. Handy for quick checks and short commands (price, start/stop bot); sensitive actions still require confirmation. |
Example conversation
Recommended flow
Finish setup
Complete AI Config (API Key, model); for group chat, set up Notification Config and enable smart reply.
Learn the entry points
Find AI Agent in the admin UI or group, try simple questions (“BTC price?”, “List my strategies”).
Complete a first task
Start with creating a strategy, running a backtest, or checking a bot—describe the goal in natural language and watch how AI Agent interprets and executes.
Go deeper
Use Strategy basics and Backtest basics alongside AI Agent for richer analysis and optimization.
Prerequisites
- Admin UI: AI Config done (provider, API Key, model).
- Group chat: besides AI config, configure the channel in Notification Config and enable smart reply.
Safety boundaries
- Two-layer architecture: AI understands intent and proposes actions; orders and config changes are executed by the trading engine under risk and permission rules—even if AI misreads intent, unsafe actions are blocked.
- No return guarantee: AI Agent provides analysis, suggestions, and operational help only; it does not guarantee returns; before live trading you’ll see risk prompts and guidance to backtest or start small.
- Permission control: What AI Agent can do is limited by system roles; if an action is unavailable, grant the right permissions in account/role management.
Next steps
- AI Config — API Key, model, and related settings
- 5-minute quick start — First strategy and backtest with AI Agent
- Create Strategy, Run Backtest — Core workflow