Everything you need to trust your AI
One platform for monitoring, explainability, fairness, and governance across traditional ML and large language models — now with one-step model connections over MCP.
Connect Any Model In Minutes, Not Sprints
The WhiteBoxXAI MCP server turns model onboarding into a single configuration block. Wire up your assistants, agent frameworks, and serving environments over the Model Context Protocol — and start monitoring, explaining, and governing them the moment they connect.
{
"mcpServers": {
"whiteboxxai": {
"command": "npx",
"args": ["-y", "@whiteboxxai/mcp-server"],
"env": { "WHITEBOXXAI_API_KEY": "wbx_..." }
}
}
}Rolling out now — MCP support ships alongside the existing Python SDK and REST API, which remain fully supported.
Connect In One Step
Point the MCP server at WhiteBoxXAI once. Your models, agents, and assistants register themselves — no glue code, no bespoke client per framework.
Standard Tooling
Model Context Protocol is the open standard for connecting AI systems to tools and data. Any MCP-capable client speaks it out of the box.
Observability Built In
Every call through the server is logged, profiled, and explainable — drift tests, token and cost tracking, and SHAP/LIME explanations without extra instrumentation.
Governed By Default
Scoped credentials, full audit trails, and policy enforcement travel with the connection, so agent access stays inside your governance boundary.
MCP ServerNew
Connect models, agents, and assistants over the Model Context Protocol with a single config block — instant monitoring, explainability, and governance, no glue code.
Real-Time Monitoring & Drift
Continuous performance tracking with Kolmogorov-Smirnov, PSI, Chi-squared, and ADWIN tests to catch data and concept drift before it costs you.
Explainability (XAI)
SHAP and LIME across tree, linear, deep, and black-box models — waterfall, force, and summary visualizations for every prediction.
Bias & Fairness Auditing
Demographic parity, equal opportunity, equalized odds, and disparate impact — with automated audits and mitigation recommendations.
LLM & RAG Observability
Track tokens, latency, and cost across providers; score toxicity and PII; evaluate retrieval quality, relevance, and faithfulness.
Intelligent Alerting
Threshold and anomaly alerts routed to Slack, email, webhooks, SMS, and PagerDuty — with correlation and fatigue reduction built in.
Compliance & Governance
ISO 42001, GDPR, CCPA, and EU AI Act mapping, model cards, audit trails, and multi-party governance review boards.
Unified Trust ScoreNew
A single 0–100 executive score per model, aggregating bias audit results, drift severity, and explainability coverage — with a full component breakdown, portfolio rollup, historical trend, and configurable weighting, included in every PDF evidence export.
AI Risk RegisterNew
Catalog, score by likelihood × impact, and track mitigation status for risk across your entire model portfolio — with a portfolio heat map, full audit trail, and entries auto-drafted from failed bias audits and high-severity drift.
RACI DashboardNew
Assign Responsible, Accountable, Consulted, and Informed roles per governance review request, drawn from your review board's real membership — with a matrix view across every board.
Lightweight SDK
Drop-in Python SDK with integrations for scikit-learn, PyTorch, TensorFlow, Hugging Face, LangChain, and XGBoost — minimal overhead.
GitHub Integration
Link production behavior back to the exact commit. Auto-register model versions and track changes across your Git history.
Fits into the tools you already use
Connect version control, alerting, and governance tooling — no rip-and-replace required.
Link production behavior back to the exact commit and auto-register model versions from your repos.
Route drift, bias, and performance alerts straight to the channels your team already watches.
Escalate critical alerts to on-call engineers with correlation and fatigue reduction built in.
Send alert notifications and scheduled compliance reports directly to stakeholder inboxes.
Pipe any alert or event into your own systems — ticketing, ChatOps, or internal automations.
Reach on-call responders by text for the alerts that can't wait for email or Slack.
Sync bias/fairness reports, drift logs, and audit trails directly into your OneTrust GRC workflows.
Feed AI governance evidence — bias audits, drift logs, and model documentation — into your Vanta compliance automation.
Push continuous AI monitoring evidence into Drata to keep your controls audit-ready without manual collection.
Don't Just Deploy AI—Audit AI
An observability and explainability platform that demystifies your production AI models — revealing why they make the decisions they do.
Build Trust
Eliminate the "black box" problem. Provide transparent, understandable explanations to business stakeholders, customers, and regulators.
Mitigate Risk
Proactively detect model decay, data quality issues, and algorithmic bias before they lead to poor business decisions or reputational damage.
Ensure Compliance
Simplify the process of auditing AI systems and demonstrating compliance with internal governance policies and external regulations.
Accelerate Debugging
Radically shorten the time it takes to diagnose and resolve production model issues with detailed root-cause analysis dashboards.
AI Risk Officers: Get professional AI bias/fairness reports, drift logs, and audit trails you can hand to an auditor without translating the technical output yourself.
See what your AI has been hiding.
Request an enterprise demo and walk through real drift detection, explainability, and governance workflows.