ADRs4AI

ADRs4AI is built around a simple observation: when a person and an AI work through a problem together, the reasoning is often more valuable than the output, yet it disappears once the session ends. The project turns that reasoning into durable Architecture Decision Records stored alongside your code, so the “why” behind each choice remains available for future sessions, teammates, and agents.

The core offering is a repository starter containing ADR and seed templates, a set of directives, an inbox protocol, and task-runner recipes. You can adopt the whole structure or pick individual features, and it is designed to be useful from the very first seed record. Records follow a consistent format with statuses such as open and answered, making them machine-readable as well as human-readable.

Supporting tools extend the method beyond the repo. A VS Code extension provides tree views of open questions, diagnostics that flag malformed records as you type, deep links, and an MCP server that lets AI agents query the record directly; it is also published on Open VSX. An iOS client lets you import a repository, review open questions, and accept answers with one tap, landing them as real commits.

This setup fits developers, technical leads, and small product teams who already use AI assistants for design and planning work and want an auditable trail without adding a separate documentation platform. It assumes familiarity with Git and a willingness to maintain records as part of normal