newcode 0.2.6

AI-powered code generation agent platform

NewCode

NewCode is an open-source, terminal-first AI coding agent for generating, editing, and reviewing code directly from your CLI.

It is maintained as a fork of code-puppy, with a more professional default UX and workflow-focused agent setup.

Why NewCode

NewCode is designed for practical software workflows in the terminal: inspect files, make edits, run commands, and iterate quickly with AI assistance.

Features

• Multi-provider model support (including OpenAI, Anthropic, Gemini, Cerebras, and compatible OpenAI-style endpoints)
• Terminal-first coding workflow (read/search/edit/delete files and run commands)
• Built-in agent system with a default coding agent and optional specialist workflows
• MCP (Model Context Protocol) integration and server management
• Interactive command-driven UX for model, agent, and settings management
• Session autosave/restore utilities
• Plugin/callback extensibility hooks

Quick start

1) Install

pip install newcode

Or with uv:

uv pip install newcode

2) Launch

newcode

Short alias:

nc

On first run, NewCode starts onboarding to help configure API keys and defaults.

3) First useful commands

/config         # Show current configuration
/model          # Select or switch model
/agent          # Select or switch agent
/colors         # Customize terminal UI colors
/api            # Manage built-in API server (start|stop|status)

Requirements

• Python >=3.11,<3.14
• At least one provider/API key for a supported model backend

Configuration

NewCode uses a legacy-compatible config filename: puppy.cfg.

• Default path: ~/.newcode/puppy.cfg
• With XDG variables: $XDG_CONFIG_HOME/newcode/puppy.cfg

Development

git clone https://github.com/janfeddersen-wq/new_code.git
cd new_code
uv pip install -e ".[dev]"

Testing

Run the test suite:

uv run pytest tests/ -v

Run linting and formatting checks:

ruff check .
ruff format --check .

Contributing

Contributions are welcome.

• Open an issue to discuss bugs, ideas, or improvements.
• Submit a pull request with a clear scope and rationale.
• Keep changes focused and include tests when applicable.

License

MIT — see LICENSE.