VT Code

cargo install vtcode
or brew install vinhnx/tap/vtcode

VT Code is a Rust-based terminal coding agent with modular architecture supporting multiple LLM providers.

Built for developers who demand precision, security, and efficiency in their coding workflows.

---

Quickstart

Installing and running VT Code

Install globally with Cargo:

cargo install vtcode

Alternatively, with Homebrew:

brew install vinhnx/tap/vtcode

Then simply run vtcode to get started:

vtcode

You can also download pre-built binaries from GitHub Releases.

Available for:

• macOS: Apple Silicon (aarch64-apple-darwin) and Intel (x86_64-apple-darwin)
• Linux: x86_64 and ARM64 architectures
• Windows: x86_64 architecture

Each archive contains the executable - extract and rename to vtcode if needed.

Configuration

Set your API key for your preferred provider:

export GEMINI_API_KEY="your_key_here"
# or
export OPENAI_API_KEY="your_key_here"
# or
export ANTHROPIC_API_KEY="your_key_here"
# or
export OPENROUTER_API_KEY="your_key_here"

Alternatively, create a .env file in your project directory:

# .env file
GEMINI_API_KEY=your_gemini_key_here
OPENAI_API_KEY=your_openai_key_here
ANTHROPIC_API_KEY=your_anthropic_key_here
OPENROUTER_API_KEY=your_openrouter_key_here

Automatic API Key Inference: VTCode automatically uses the correct environment variable based on your provider setting in vtcode.toml:

• provider = "openai" → OPENAI_API_KEY
• provider = "anthropic" → ANTHROPIC_API_KEY
• provider = "gemini" → GEMINI_API_KEY
• provider = "deepseek" → DEEPSEEK_API_KEY
• provider = "openrouter" → OPENROUTER_API_KEY

VT Code supports advanced configuration via vtcode.toml. See Configuration for details.

Using OpenRouter models

OpenRouter support unlocks any hosted model by ID, including the latest Grok and Qwen3 coding releases.

vtcode --provider openrouter --model x-ai/grok-code-fast-1 chat

Or persist the configuration in vtcode.toml:

[agent]
provider = "openrouter"
default_model = "qwen/qwen3-coder"

Custom model IDs are accepted as long as they match your OpenRouter account access. Streaming and tool-calling work out of the box using the OpenAI-compatible Responses API.

---

Core Features

Multi-Provider AI Support

• Gemini, OpenAI, Anthropic, OpenRouter, and DeepSeek integration
• Automatic provider selection and failover
• Cost optimization with safety controls

Advanced Code Intelligence

• Tree-sitter parsing for 6+ languages (Rust, Python, JavaScript, TypeScript, Go, Java)
• Semantic code analysis and pattern recognition
• Intelligent refactoring and optimization suggestions

Enterprise Security

• Workspace boundary enforcement
• Configurable command allowlists
• Human-in-the-loop controls for safety
• Comprehensive audit logging

Modular Architecture

• Trait-based tool system for extensibility
• Multi-mode execution (terminal, PTY, streaming)
• Intelligent caching and performance optimization
• Plugin architecture for custom tools

---

Documentation

• Getting Started - Installation and basic usage
• Configuration - Advanced configuration options
• Architecture - Technical architecture details
• Advanced Features - Safety controls and debug mode
• API Reference - Complete API documentation
• Contributing - Development guidelines

---

License

This project is licensed under the MIT License - see LICENSE for details.