What it does
The server exposes WhiteBoxXAI capabilities through the Model Context Protocol: a model registry, Git-backed model diffing, prediction logging, drift detection, bias and fairness auditing, ISO/IEC 42001 readiness and compliance evidence, AI Risk Register and governance review queries, prompt governance, asynchronous SHAP and LIME explanations, LLM explainability, LLM call monitoring, RAG pipeline monitoring, content safety analysis, alert management, multi-agent workflow tracking and model performance metrics. It provides 197 tools across 18 domains and can display sandboxed interactive audit cards through MCP Apps.
It is intended for teams that want an MCP-compatible client, such as Claude Desktop, Claude Code, LangChain or a custom agent harness, to query and act on WhiteBoxXAI data. The publisher also provides agent skills that help a coding agent set up the server and call its tools correctly. Runnable examples cover LangChain, Semantic Kernel and a raw MCP client.
Local use runs over stdio, with the API key supplied through an environment variable and the server started from a console script. The hosted Streamable HTTP endpoint accepts an API key or personal JWT in a Bearer authorization header. Releases are published through GitHub releases.
Limitations and requirements
The stdio credential is a scoped, revocable API key and is not tied to a human account's password or session lifecycle. Support for the hosted HTTP endpoint follows the caller's underlying Cloud plan, and self-hosted stdio use is supported on a best-effort basis. The tools whitebox_agent_workflows_cost_breakdown, _bottlenecks and _timeline can take up to about 30 seconds. The publisher lists further known v1 limitations in its integration documentation.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 13 Nov 2026 · its page on the source
WhiteBoxXAI
AI observability, explainability and governance: drift, fairness, ISO 42001, risk register.
v0.3.2As shown on registry.modelcontextprotocol.io when we last read it