code-ant

Autonomous code-improvement agent for the command line.

code-ant crawls your git repository, suggests safe micro-improvements (unused imports, missing docstrings, type hints, lint fixes, antipatterns), and commits each change individually — validated by your test suite.

Installation

Homebrew

brew tap helloodokai/code-ant
brew install code-ant

From source

cargo build --release

Configuration

code-ant reads configuration from two optional TOML files:

1. User-level: ~/.code-ant/config.toml
2. Project-level: ./.code-ant/config.toml

Project-level config overrides user-level config. CLI flags and environment variables override both.

Example config.toml

[provider]
default = "anthropic"

[provider.anthropic]
api_key = "sk-ant-api03-..."
model = "claude-sonnet-4-6"

[provider.openai]
api_key = "sk-proj-..."
model = "gpt-4"

[provider.ollama_local]
host = "http://localhost:11434"
model = "qwen2.5-coder:32b"

Environment variables (highest priority)

• CODE_ANT_PROVIDER
• CODE_ANT_MODEL
• CODE_ANT_API_KEY

Usage

# Run with defaults (auto-detects test suite)
code-ant

# Limit to 5 commits
code-ant --max-commits 5

# Only Python files, skip test verification
code-ant --include "**/*.py" --skip-tests

# Custom model provider
code-ant --provider anthropic --model claude-sonnet-4-6 --api-key $ANTHROPIC_API_KEY

Environment variables

• CODE_ANT_PROVIDER — override provider (ollama_cloud, ollama_local, openai, anthropic)
• CODE_ANT_MODEL — override model name
• CODE_ANT_API_KEY — override API key

Supported model providers

• Ollama Cloud (default)
• Ollama Local
• OpenAI
• Anthropic

License

MIT