DecisionKit

•
8
•

Deliver fast, cost-effective, probability-scored decisions for AI coding agents, automation scripts, and CI pipelines.

DecisionKit provides provider-neutral, closed-form decision-making for AI agents, developer scripts, and CI environments. Instead of generating verbose natural language responses, it queries specialized decision APIs—including OpenAI Decisions, TypeSafe AI's Jev, and Claude Code—to return exact probability distributions for binary questions, multiple-choice selections, ratings, and candidate rankings. Operating as a CLI, Claude Code plugin, or standard Model Context Protocol (MCP) server, it enables programmatic workflows to execute instant, low-latency classifications and triage tasks with strict confidence thresholding.

主要功能

01Supports multiple question types including binary true/false, multiple choice, scalar scoring, and candidate ranking
02Provides provider-neutral integration supporting OpenAI Decisions, Jev (TypeSafe AI / OpenRouter), and Claude Code
03Outputs calibrated probability scores per answer rather than verbose natural language text
04Follows grep-style exit codes and customizable thresholds for seamless shell script and CI/CD integration
05Offers multi-interface support as a CLI tool, Claude Code plugin, or universal MCP server
068 GitHub stars

使用案例

01Automating issue and bug triage in CI/CD pipelines with configurable confidence thresholds
02Evaluating Git diffs programmatically in deployment scripts to detect breaking API changes
03Assisting coding agents in ranking relevant code files or candidate solutions during automated debugging