AGENT BUILDER

From English to Autonomous Agent

Describe your strategy in plain language. YorN generates inspectable code, validates it, and lets you test before deployment.

Describe
Generate
Validate
Deploy

Your Words Become Working Code

Start with a natural-language description of your strategy. YorN parses your intent, maps it to market data sources and signal logic, then generates a complete Python strategy class that plugs directly into the execution engine.

The generation pipeline handles everything: importing the right market data feeds, building evaluation logic, sizing positions with Kelly criterion, and wiring up risk management -- all from a single prompt.

  • Supports any prediction market category
  • Automatically selects relevant data sources
  • Generates position sizing and risk limits
  • Full code review available before deployment
YOUR PROMPT
"Buy YES on any presidential swing-state contract when the 7-day polling average shows a 3+ point shift, but limit position size to 5% of bankroll and exit if momentum reverses."
generated_strategy.py 247 LINES
# Auto-generated by YorN Agent Builder
# Strategy: Swing State Momentum

class SwingStateMomentum(BaseAgent):
    name = "swing_state_momentum"
    version = "1.0.0"
    category = "politics"

    def configure(self):
        self.lookback = 7 # days
        self.threshold = 3.0 # points
        self.max_position = 0.05

    def evaluate(self, market):
        polls = self.fetch_polls(
            state=market.state,
            days=self.lookback
        )
        shift = self.momentum(polls)

        if shift >= self.threshold:
            return Signal(
                side="YES",
                size=self.kelly(shift),
                confidence=shift / 10
            )
AST VALIDATION TREE
Module SwingStateMomentum PASS
ClassDef inherits BaseAgent PASS
FunctionDef configure() PASS
FunctionDef evaluate() PASS
Call fetch_polls() PASS
Call momentum() PASS
Return Signal(...) PASS
Import os.system BLOCKED
Call eval() BLOCKED
7 nodes validated 2 nodes blocked Safety: PASS

AST Safety Validation

Every generated strategy passes through a rigorous Abstract Syntax Tree validation layer before it can execute. The AST walker inspects every node of the generated code, verifying that it only calls approved functions, accesses approved modules, and stays within the sandboxed execution environment.

Dangerous operations like file system access, network calls outside the approved API list, dynamic code evaluation, and process spawning are automatically detected and blocked. You get the power of AI-generated code with the safety of a locked-down sandbox.

  • Every AST node inspected before execution
  • Blocked: eval, exec, os, subprocess, network
  • Allowlisted API calls only
  • Transparent validation report for every build

Deploy to Paper Without Build Work

Once validated, your agent can enter paper-trading mode from the product UI. No build pipelines, no infrastructure provisioning, no DevOps tickets. Click a button and your agent starts scanning markets and generating signals.

Every agent can start with simulated funds. Set your own review criteria -- minimum number of trades, drawdown limits, or metric thresholds -- then decide when live trading should be enabled on a supported paid tier.

  • Instant paper-trading deployment
  • Custom graduation thresholds for live trading
  • Automatic or manual promotion pipeline
  • Real-time performance monitoring from day one
Agent: SwingStateMomentum
Status PAPER TRADING
Mode Autonomous
Paper Balance $10,247.83
Trades Executed 34 / 50 required
Win Rate 68.2%
Sharpe Ratio 1.84
Graduation 16 trades remaining
Risk Limit 5% max position
Daily Loss Limit $500 (not triggered)
247
Avg lines generated per agent
Custom strategies
<10s
Generation time

Strategy Lab: Describe It, AI Builds It

The Strategy Lab is a conversational AI workspace with 9 specialized tools. Describe your trading idea in plain English and the AI generates a complete Python DirectionalStrategy subclass with signal logic, position sizing, and risk rules. Backtest it instantly, run crowd simulations, compare strategies — then deploy as a live agent directly from the chat.

Momentum

Poll-driven momentum signals

Sentiment

News and social sentiment analysis

Mean Reversion

Contracts that overshoot fair value

Arbitrage

Cross-market price discrepancies

Custom

Build anything from scratch

Contrarian

Fade extreme market positions

A Conversational AI Workspace

The Strategy Lab is more than a chat. It's an AI workspace with 9 specialized tools that the AI can call autonomously during your conversation — backtesting strategies, running crowd simulations, generating Python code, comparing performance, and deploying agents. All from a single chat interface.

  • generate_strategy_code — AI writes a full Python DirectionalStrategy subclass
  • backtest_custom_code — Run the generated code through the backtest engine instantly
  • run_crowd_simulation — Simulate 10K market participants on any event
  • deploy_as_agent — Turn an approved strategy into a managed agent
  • compare_strategies — Head-to-head performance comparison with metrics
strategy_lab_chat STREAMING
You: Build me a momentum strategy for NBA games
    that buys underdogs when they're down 10+ points

[tool_call] generate_strategy_code
  Writing DirectionalStrategy subclass...

[tool_call] backtest_custom_code
  Sharpe: 1.42 | Win Rate: 63% | Return: +18.5%

AI: Strategy backtested successfully. 63% win rate
    with a 1.42 Sharpe. Want me to deploy it?

[tool_call] deploy_as_agent
Agent "NBA Underdog Momentum" deployed.

Describe a Strategy. Deploy an Agent.

No coding skills required. Your first agent is free to build and paper trade.