ADRs4AI

What is ADRs4AI?

ADRs4AI is a method and toolkit for human-AI collaboration. It captures unresolved questions raised by AI agents and records human decisions as version-controlled files. Every decision includes options, a recommendation, rationale, confidence level, and a falsifier condition. When a human accepts or overrides an answer, it lands as a git commit. This keeps the reasoning behind choices intact so it compounds across projects.

The package includes a starter template, a VS Code extension, an iOS client, a CLI, and an MCP server. The extension provides tree views of open questions, syntax diagnostics, and deep links. The iOS app lets users review questions and tap accept to commit answers. The MCP server and CLI allow other tools and agents to query and update the record.

The workflow starts by cloning the template and initializing a project file. Users dump their initial thoughts, recruit an AI agent, and answer the first question it raises. Each open question is searchable via grep or surfaced through the editor and mobile app. The system is designed for both experienced engineers and beginners who want to preserve decision context without relying on vector databases or transcript archives.

ADRs4AI features

• AI raises unresolved questions using ORRCF format
• Human rulings land directly as git commits
• VS Code extension with tree views and diagnostics
• iOS client for reviewing and answering on mobile
• MCP server and CLI for agent integration
• Claude Code plugin for automated workflow
• Public template repository with full methodology

Who uses ADRs4AI and for what

• Tracking architectural decisions alongside code in git
• Capturing AI reasoning and recommendations before implementation
• Reviewing project questions during commutes via iOS app
• Integrating decision records into CI/CD with GitHub Actions
• Maintaining audit trails for compliance or team handoffs

ADRs4AI FAQ

Comments (0)

Comments and upvotes open when this product launches on ProductPerch.