Drex Decision

Drex Decision gives AI agents a dedicated decision layer for situations where they need to choose between a defined set of alternatives.

Instead of relying on an unstructured opinion, Drex accepts a question, 2–12 explicit choices, and optional supporting evidence. It returns a selected choice together with a probability distribution across the available alternatives.

Drex is designed for agent workflows that need a clear, machine-readable decision before taking the next action.

What Drex does

• Accepts a bounded decision question
• Evaluates 2–12 explicit alternatives
• Accepts optional evidence or context
• Returns the selected choice
• Returns probabilities for every alternative
• Provides structured output suitable for downstream automation

Example uses

• Choosing between implementation approaches
• Selecting which task or issue to prioritize
• Comparing candidate actions
• Routing an agent workflow
• Evaluating competing hypotheses
• Making structured go/no-go or multi-option decisions

Agent integration

The skill connects to Drex through its MCP server and exposes the drex_decide tool.

Configuration is provided for MCP-capable clients including Claude Code, Cursor, Codex CLI, GitHub Copilot/VS Code, Hermes, and other compatible MCP clients.

OpenWebUI can connect through an MCP-to-OpenAPI bridge such as mcpo.

Requirements

• Python 3.10+
• A compatible MCP client
• A Drex API key
• Network access to the hosted Drex service

Drex API/service access is separately controlled and is not granted merely by possession of the skill package.