Keep the why attached to the code.

repowise mines architectural decisions from eight sources, not just an ADR folder: ADR files, CHANGELOGs, PR bodies, inline markers in code, git archaeology, READMEs, comment archaeology, and agent transcripts. The result is a single decision layer that captures rationale wherever your team actually recorded it, even when they never wrote a formal record.

Every decision carries a verification status: verified, fuzzy, or unverified. A verified decision is backed by concrete evidence spans, the exact passages in ADRs, commit messages, PR bodies, or comments that support it, so you can read the source and trust the claim. Fuzzy and unverified decisions are surfaced honestly instead of presented as fact, so nothing is invented without provenance.

Decisions rarely stand alone. repowise connects them with supersedes, refines, and conflicts_with edges into a decision graph, so you can follow a chain from the original choice through every later revision. When a newer decision supersedes an older one, the lineage records it, and contradictory decisions are flagged instead of silently coexisting.

A decision is only as current as the code it governs. Each decision is linked to graph nodes, the files and modules it applies to, and it ages when those files get commits. As the governed code drifts away from the recorded rationale, repowise marks the decision as stale so you know when the why no longer matches the what.

get_why is the MCP tool that returns architectural decision rationale in a single call: the decision, its status, the evidence spans behind it, and its supersession lineage chain. It is the tool an AI agent or a reviewer calls to answer why is the code shaped this way before a refactor. When no ADRs exist for a file, get_why falls back to git archaeology so you still get the most likely rationale instead of an empty answer.

A folder of ADR markdown files only captures the decisions someone remembered to write down, and it never tells you when one goes stale. repowise mines eight sources, attaches a verification status and evidence spans, links each decision to the graph, tracks staleness as the code changes, connects decisions by lineage, and surfaces governance risk in PR reviews. It is the difference between a static archive and a living decision layer.

Decisions are governance-aware. They surface in PR risk reviews so reviewers see when a change touches code governed by an existing decision, when a hotspot has no decision governing it at all, and when two decisions contradict each other. The Repowise PR bot can flag these ungoverned hotspots and contradictory decisions deterministically, with no LLM in the scoring path.

Yes. repowise is open source under AGPL-3.0, so every step of how decisions are mined, verified, linked, and aged is inspectable and reproducible on your own repo. The same decision layer is available self-hosted and on the paid hosted tier.