Skip to main content

​
Example: AI trading agent deciding BUY / SELL

An AI agent monitors markets and decides whether to BUY, SELL, or HOLD an asset.

​
Flow

  1. The agent fetches external data (prices, indicators, market data APIs)
  2. The agent sends a prompt to an AI model (e.g. “Given this data, should I BUY or SELL?”)
  3. The AI model returns a response (e.g. BUY with a confidence score)
  4. The agent executes the trade

​
Where Prado is used

  • HTTP runtime
    • Records the external data APIs the agent called
    • Records the AI model request and response (or hashes of them)
  • Optional JS / Python runtime
    • Records deterministic post-processing logic
    • Example: reject low confidence, cap position size

​
What Prado proves

  • What market data the agent observed
  • What prompt was sent to the AI model (or its hash)
  • What the AI model returned
  • Whether deterministic logic modified the result
  • That this evidence was not altered later

​
What Prado does NOT prove

  • That the trade was correct
  • That the model reasoning was sound
  • That the outcome was profitable

​
Why this matters

When AI agents make financial decisions, failures are inevitable. Prado allows teams to inspect and audit decisions after the fact, instead of relying on logs or explanations.