comparisongit-agent vs aicommit2
aicommit2 is an AI commit message generator that supports multiple LLM providers and suggests messages for your currently staged diff; unlike git-agent, it does not own the full Git handoff.
feature comparison
| Feature | git-agent | aicommit2 |
|---|---|---|
| Atomic commit splitting | git-agent — Yes | aicommit2 — No |
| Conventional Commits format | git-agent — Yes | aicommit2 — Yes |
| LLM-backed message drafting | git-agent — Yes | aicommit2 — Yes |
| Pre-commit hook integration with retry loop | git-agent — Yes | aicommit2 — No |
| Dry-run preview before committing | git-agent — Yes | aicommit2 — No |
| Amend last commit message | git-agent — Yes | aicommit2 — No |
| Free shared gateway | git-agent — Yes | aicommit2no — requires your own API key |
| Config-file based scope definitions | git-agentyes — .git-agent/config.yml | aicommit2limited |
install
git-agent
brew install gitagenthq/brew/git-agentaicommit2
npm install -g aicommit2output example
git-agent
feat(auth): add refresh token rotation on silent renew
- generate a new refresh token on every silent renew to limit token reuse window
- invalidate the previous refresh token immediately after issuing the replacement
- store the new token in the same secure HttpOnly cookie
Long-lived refresh tokens were reusable indefinitely; rotation limits the
damage window to a single reuse attempt per token.aicommit2
feat: add token refresh rotationFAQ
Does aicommit2 split staged changes into multiple commits?No. aicommit2 generates a single commit message for whatever is staged. git-agent's atomic split planning groups staged hunks into logically separate commits before writing any of them.
Do I need an API key to use git-agent?No. Official release binaries use a free shared gateway with zero config — just hand your work to
git-agent --intent "...". You can optionally bring your own key and point it at your own OpenAI, Gemini, or Claude endpoint via config.Can git-agent replace aicommit2 with no workflow changes?Yes. Run
git-agent --intent "..." after completing the work. The main addition is autonomous atomic splitting: git-agent discovers the working tree, stages logical groups, and commits them independently.