Intelligence · AI Agent

An AI strategy builder, checked by the engine.

Runbot's AI Agent is in-house intelligence built for trading and finance. A trader describes a strategy in plain English; the agent writes its rules one by one, checks each against the engine, and runs a first backtest before it answers.

See how it works
200+indicators and trading rules, plus unlimited custom ones
45ready-made templates
7steps in the Visual Builder
15M+strategies created
01

AI Agent

Turns a sentence into rules the engine accepts, and draws on tens of thousands of strategy patterns that passed real backtests.

02

Every rule checked

Each building block is checked against the engine before the strategy is saved.

03

A first backtest

The new strategy is tested before the answer arrives, so the reply already has numbers in it.

04

Visual Builder

For users who prefer to click: seven steps, every indicator and rule, then the agent builds exactly what was picked.

05

Templates

45 ready-made strategies, from a simple RSI to full trading systems, adapted by the agent to your market.

06

Custom indicators

Describe a missing indicator or trading rule in a prompt; the AI builds it, checks it on historical data and adds it to the engine.

From one sentence to a strategy

  1. 01

    Describe the idea. The market, the indicators, the timeframe and the risk you accept, for example: "A BTC scalper on 5-minute candles with RSI and Bollinger Bands." The agent asks when something is missing.

  2. 02

    The agent picks the building blocks. Indicators, entry rules, exits and sizing, each set one by one.

  3. 03

    Every rule is checked. Each step shows in the chat, and every rule is checked against the engine before the strategy is saved.

  4. 04

    A first backtest runs. On real order books, before the answer arrives.

  5. 05

    The agent explains. The rules it built, the risk it takes and how the first backtest went, then what to do next.

What a strategy is made of

Every strategy is a set of rules the engine can run. Eight blocks hold the market, selectors, indicators, sizing, entries, order methods, exits and adjustments: indicators raise signals, entry rules turn them into trades, exits and sizing manage them.

  • 150 indicator conditions, 32 exit rules, 16 sizing methods and 10 order methods.
  • Trailing stops, break-evens, partial take-profits and mid-trade rules.
  • The engine checks the rules candle by candle, the same way in backtests and live bots.

The anatomy of a strategy and the indicator catalog are in the docs.

Keep improving it in the same conversation

Ask for another timeframe, a tighter stop or an extra filter: each change is saved as a new version you can compare or restore. The AI analyst reads every backtest and every trade, names the biggest problems and proposes a fix it can apply, and the AI Optimizer searches for better settings in the background.

For three market families

The same agent builds strategies for perpetual futures on Hyperliquid, Up/Down markets on Polymarket and pari-mutuel contests on glgl.fun. It also serves white-label partners under their brand, and AI agents over the MCP API.

FAQ

Frequently asked questions.

Do I need to code to build a trading strategy with Runbot?

No. Describe the strategy in plain English to the AI Agent, or build it in the Visual Builder.

How does the AI avoid building a strategy the engine cannot run?

It writes the rules one by one and checks each against the engine before the strategy is saved, then runs a first backtest before it answers.

Can my own AI agent use the strategy builder?

Yes. Over the MCP API, paid per call over x402, or by connecting to your Runbot account over MCP or A2A: see Bring your own agent.

Does a good backtest mean the strategy will make money?

No. Backtest results are hypothetical and do not guarantee future performance. Read the risk disclaimer.

Let's talk

The whole stack, under your brand.

Offer the AI Agent, the backtester and the bots to your users under your own name. We will show you the app under your brand.