Set up
Runs on your computer. Pick your agent and operating system.
Before you start
- uv (`curl -LsSf https://astral.sh/uv/install.sh | sh`)
{
"mcpServers": {
"redwood": {
"command": "uvx",
"args": [
"redwood"
]
}
}
}same format Cursor, GitHub Copilot CLI, Junie, JetBrains AI Assistant, Augment (Auggie CLI), Factory Droid, Kiro, Devin CLI +12
Values in angle brackets are placeholders. Restart Claude Desktop after editing its config.Claude Desktop docs
README
synced Oct 21, 2026🌲 Redwood
Redwood is a powerful, highly-configurable CLI and GUI client for Google's Gemini models, built with a focus on transparency and user control. It integrates deeply with the Model Context Protocol (MCP) to extend its capabilities with custom tools and data sources.
✨ Features
- 🖥️ Dual Interface: Choice between a sleek Flet GUI and a traditional terminal CLI.
- 🛠️ MCP Native: Seamlessly use MCP tools for tasks like web scraping, file management, and more.
- 🎨 Image Generation: Built-in support for image generation models via MCP.
- 🧠 Full Prompt Control: Complete control over the system prompt via a single configuration file.
- 📝 Skills Integration: Ability to read and utilize specialized agent skills.
- 🗄️ Persistent Storage: Integrated database for storing user-defined data and lists.
🚀 Quick Start
1. Prerequisites
Ensure you have uv installed:
Mac Users:
brew install uv
Linux users:
curl -LsSf https://astral.sh/uv/install.sh | sh
2. Launching Redwood
You can run Redwood directly without cloning the repository:
Run the GUI:
uvx --from git+https://github.com/douglas-gibbons/redwood gui
...or run the CLI:
uvx --from git+https://github.com/douglas-gibbons/redwood cli
...or run the Web GUI:
uvx --from git+https://github.com/douglas-gibbons/redwood web
This will start a local web server making the GUI accessible through your browser at localhost:8550.
3. Configuration
On the first run, Redwood creates a configuration file at ~/.config/redwood/redwood.yaml and prompts you for a Gemini API key which it adds to the file.
You can configure any aspect of Redwood in this file, or ask Gemini (through Redwood) to do it for you.
🛠️ Included Tools
Redwood comes bundled with several core MCP capabilities:
- Image Generation: Saves generated images to a local directory for easy access.
- Agent: A sub-agent tool allowing for parallel task management.
- General Utilities:
- Time: Provides current date and time context.
- Shell: Securely run local commands.
- Web Scraper: Extract content from the web.
- Skills Reader: Loads markdown-based skills for specific workflows.
- Database: A simple key-value store. You can prompt the model to use this for persistent lists (e.g., "Add milk to my shopping list").
💡 Pro Tips
Set Working Directory
The model doesn't automatically know your current directory. Use the /location command in the CLI to set your working directory. This injects the current path into the prompt, enabling relative file operations.
Visualizing Tools
The GUI features a dedicated pane to monitor MCP tool interactions in real-time, providing transparency into the agent's thought process.
🛠️ Development
Requirements
- Python with
uv(brew install uv) make(brew install make)
Local Setup
git clone https://github.com/douglas-gibbons/redwood.git
cd redwood
Useful Commands
- Run CLI:
make cli - Run GUI:
make gui - Run Web GUI:
make web - Run Server:
make server - Run Tests:
make test - Inspect MCP:
npx @modelcontextprotocol/inspector make server
🤝 Contributing
Contributions are welcome! Please feel free to submit a Pull Request or open an issue.
- Fork the repository.
- Create your feature branch (
git checkout -b feature/amazing-feature). - Commit your changes (
git commit -m 'Add some amazing feature'). - Open a Pull Request.
Created by Douglas Gibbons
Disclaimer: This is not an official Google project. It is an open-source project maintained by Douglas Gibbons.