Kairos is an open-source, real-time risk and fraud decisioning platform designed to evaluate sign-ups, logins, orders, payments, and payouts instantly. Built as an alternative to proprietary solutions like Sift and Sardine, it addresses two major bottlenecks in fraud prevention: the scarcity of labeled training data and the rapid adaptation of fraudsters.

Just like developers explore graphics tech through projects such as Ultra-Water: Renderizador Oceánico FFT en Tiempo Real para Three.js, engineering teams are now looking at open-source AI infrastructure to secure their applications against sophisticated attacks and automated fraud vectors.

How do decision models and traditional ML work together?

The platform combines Finic's decision model, Grayson, with traditional machine learning algorithms like logistic regression and gradient-boosted trees. Grayson answers plain-language questions about requests—such as checking for throwaway emails or reshippers—operating effectively even on day one without historical chargeback data.

Traditional models precisely learn from internal metrics, and Kairos merges both approaches so that Grayson's probabilistic answers serve as features alongside standard numerical fields during automated model experiments.

How do AI agents tune and manage workflows?

Kairos features an Model Context Protocol (MCP) server that lets coding agents like Claude Code, Cursor, or Codex act as fraud machine learning engineers. An agent can independently build workflows, propose model variants, test them against historical data, and deploy the highest-performing option.

You can add the MCP transport in your development environment using the command line tool and prompt your agent to review decisions or execute validation experiments seamlessly.

How do I install and configure the SDK?

To start using Kairos in your backend infrastructure, you can install the official TypeScript or Python client packages directly from your package manager.

Run the following command to install the TypeScript SDK in your project before initializing the client with your API credentials:

  1. Ensure your environment meets the prerequisites such as Node.js or Python.
  2. Install the client library using npm or pip.
  3. Initialize the Kairos instance with your API key and base URL.
  4. Define your business actions and track user events.