Selvedge vs. Git AI
Git AI stores attribution metadata in git notes under its own Git AI standard. Selvedge is complementary: it adds live reasoning capture, a query layer, and changesets, and can also emit the Agent Trace v0.1.0 record format.
| Selvedge | Git AI | |
|---|---|---|
| Reasoning source | Captured live, by the agent in the same context that produced the change | Attribution metadata |
| Granularity | Entity — DB column, table, env var, dep, API route, function | Line |
| Mechanism | MCP server — the agent calls it as work happens | Agent-invoked checkpoint → Git notes at commit |
| Grouping | Changesets — named feature/task slugs across many entities | None |
| Prior attempts | Yes — prior_attempts surfaces tried-and-rejected paths | None |
| Storage | SQLite, zero deps | Git notes |
Where Selvedge is different
Section titled “Where Selvedge is different”- Reasoning, not just attribution. Git AI records who/what touched a line. Selvedge records why, in the agent’s own words, captured live.
- A query layer.
selvedge blame,diff,history,changeset, andsearchare first-class. Git-notes attribution has no equivalent CLI you can pipe. - Continuous vs. cooperative. Git AI’s own README states it does not rely on git hooks: the agent calls
git-ai checkpoint, and attribution lands in git notes at commit time. That is cooperative snapshotting — it records what the agent chose to check point, when it chose to. Selvedge captures at the moment of each change. - Agent Trace interop. Selvedge also emits Agent Trace v0.1.0 records — see the Agent Trace interop page. (Git AI itself has since moved to its own Git AI standard rather than Agent Trace.)
- Entity-level + changesets for querying the history of a column or a whole feature, not a line.
When Git AI might fit you better
Section titled “When Git AI might fit you better”If all you need is line-level attribution metadata living in git notes under Git AI’s own standard — and you don’t need captured reasoning or a query layer — Git AI is a focused fit.