aicommit
A fast, opinionated CLI that generates Git commit messages and reviews staged diffs using AI. Written in Rust, installed as aic.
Quick start
# Install via Homebrew (macOS)
brew install russmckendrick/tap/aicommit
# Install via WinGet (Windows)
winget install --id RussMckendrick.Aicommit -e
# Or download a binary from GitHub Releases
# https://github.com/russmckendrick/aicommit/releases
# Run setup to configure your provider and credentials
aic setup
Usage
Stage your changes and run aic - it generates a conventional commit message, shows it for confirmation, and commits on approval:
git add -p
aic
In an interactive terminal, if files are already staged, aic lets you continue, unstage selected files, or abort before generation starts.
If nothing is staged, aic lets you stage all changed files, choose files interactively, or cancel.
With --yes, it stages all changed files automatically.
In the normal interactive flow, aic can also split one staged change set into multiple file-group commits when the work clearly spans separate concerns.
Flags
| Flag | Description |
|---|
-c, --context | Add context for the AI (e.g., -c "closes #42") |
-y, --yes | Skip commit prompts and auto-stage/auto-push when unambiguous |
-d, --dry-run | Print the message without committing |
--amend | Regenerate and amend the last commit |
--fgm | Use the full GitMoji specification |
Review staged changes
Get AI-powered feedback on your staged diff before committing - findings are grouped by severity and rendered in the terminal:
aic review
aic review -c "focus on security"
aic review inspects the current staged set as-is. In an interactive terminal, if nothing is staged, it offers the same staging menu as aic, but it does not unstage files during review.
Rewrite commit messages
Clean up the last N commit messages on your branch before opening a PR:
aic log
aic log -n 3
Draft a pull request
Generate a PR title and Markdown description from the branch commits and diff against a base branch:
aic pr
aic pr --base origin/main
aic pr -c "highlight rollout risk"
aic pr --yes
Browse history
View recent AI-generated commit messages, PR drafts, and reviews:
aic history # interactive by default in a terminal
aic history --non-interactive
aic history --all
aic history --verbose
aic history --kind review
Visualize the codebase
Generate SVG visualizations of file structure, commit history, and activity:
aic map tree # file hierarchy treemap
aic map history # zigzag commit timeline
aic map heat # file modification heatmap
aic map activity # GitHub-style contribution grid
More commands
aic setup # Interactive provider setup
aic config set KEY=value # Set configuration
aic models --refresh # List available models
aic hook set # Install prepare-commit-msg hook
aic completions zsh # Generate shell completions
Features
- Conventional commits with optional GitMoji prefixes
- Scope hints - detects likely scopes from changed file paths
- Branch context - extracts ticket/issue references from branch names
- Diff review - AI-powered code review with severity grouping
- Message rewriting - clean up commit history with
aic log
- Pull request drafts - generate a PR title and Markdown body with
aic pr
- History - browse past generated messages, PR drafts, and reviews
- Large diff handling - automatic chunking and synthesis
- Custom prompts - swap the system prompt without recompiling
- Split commits - turn one staged change set into multiple file-group commits when concerns are separate
- Visualization - generate SVG treemaps, timelines, heatmaps, and activity graphs with
aic map
- Provider choice - OpenAI, Azure OpenAI, Anthropic, Groq, Ollama, Claude Code, Codex, GitHub Copilot CLI, and custom OpenAI-compatible endpoints
Configuration
Set values globally with aic config set or per-invocation with environment variables:
aic config set AIC_MODEL=gpt-5.4-mini AIC_EMOJI=true
AIC_LANGUAGE=french aic
See docs/configuration.md for the full key reference.
Documentation
Detailed docs live in docs/:
- Installation - Homebrew, WinGet, GitHub Releases, from source
- Usage - commit workflow, review, flags, hooks
- Configuration - keys, prompt templates, ignore files
- Providers - OpenAI, Azure OpenAI, Anthropic, Groq, Ollama, Claude Code, Codex, GitHub Copilot CLI, custom endpoints
- Architecture - module layout and data flow
- Roadmap - planned and completed features
License
MIT