Skip to content

Audit evidence and explanations you don't have to translate yourself

WhiteBox XAI packages what your AI actually did — SHAP and LIME explanations, bias audits, drift history, governance decisions — into reports you can hand to an auditor or a board without building a slide deck at midnight. It's also the observability and monitoring platform your engineering team uses to produce that evidence in the first place.

Generate an Audit Report Get started Install the SDK

Start here

  • Audit & Explanation Reports

    Generate bias-audit, compliance, and explainability reports from the dashboard — no code required — with every number traceable back to real computed data.

  • Getting Started

    Create your account, install the SDK, register your first model, and see monitoring in action — end to end.

  • User Guide

    A tour of the platform: dashboards, explanations, drift detection, bias auditing, LLM monitoring, alerts, and reports.

  • Python SDK

    Instrument your models in a few lines. Decorators, async logging, offline buffering, and multi-agent monitoring included.

  • Integrations

    Native support for scikit-learn, PyTorch, TensorFlow, Hugging Face, LangChain, XGBoost, LightGBM, GitHub, n8n — and an MCP server for any AI client.

What you can do with WhiteBox XAI

  • Explain every decision

    Generate human-readable explanations with SHAP and LIME, so you can answer why a model made a given prediction — and hand the explanation to someone who wasn't in the room.

  • Audit for fairness

    Run bias and fairness audits and produce the evidence your stakeholders and regulators expect — not a summary of one.

  • Govern with confidence

    Multi-party governance review boards, automated periodic reviews, and an immutable decision archive — aligned with ISO 42001 and the EU AI Act.

  • Track AI risk

    A structured AI Risk Register with owners, likelihood × impact scoring, and a full audit trail — the artifact ISO 42001 and the EU AI Act ask for.

  • Report one number

    A Trust Score per model, combining fairness, drift, and explainability into a single 0–100 index your board can actually read.

  • Monitor in real time

    Track production models for performance degradation, data drift, and concept drift — with proactive alerts before problems reach your users.

Install the SDK

pip install whitebox-xai-sdk

from whiteboxxai import WhiteBoxXAI

client = WhiteBoxXAI(api_key="your-api-key")

Then follow the Getting Started guide to register a model and log your first predictions.

Need help?