VT Code: Open-Source Rust TUI Coding Agent with Multi-Provider Support and Agent Skills

✍️ OpenClawRadar📅 Published: April 25, 2026🔗 Source
VT Code: Open-Source Rust TUI Coding Agent with Multi-Provider Support and Agent Skills
Ad

VT Code is an open-source terminal-based AI coding agent built in Rust (using Ratatui for the TUI). It supports multiple LLM providers including Anthropic, OpenAI, Gemini, and Codex, as well as any open-source model via local inference through LM Studio and Ollama (experimental). The agent is Agent Skills, Model Context Protocol (MCP), and Agent Client Protocol (ACP) ready.

Key Features

  • Multi-provider LLM support with automatic failover and efficient context management.
  • Semantic code understanding powered by ast-grep for structured code search and ripgrep for fast grep operations.
  • Agent Skills following the open Agent Skills standard, enabling external capability extension through a standardized skill system.
  • Shell safety via process hardening (vtcode-process-hardening) and bash runner (vtcode-bash-runner).
  • Terminal detection and PTY integration (vtcode-terminal-detection, vtcode-ghostty-vt-sys) for richer screen snapshots.
  • Extensions for VSCode and Zed are included in the repo.
Ad

Installation

Native Installer (macOS/Linux):

curl -fsSL https://raw.githubusercontent.com/vinhnx/vtcode/main/scripts/install.sh | bash
# Skip ripgrep + ast-grep bundle:
curl -fsSL https://raw.githubusercontent.com/vinhnx/vtcode/main/scripts/install.sh | bash -s -- --without-search-tools

Windows (PowerShell):

irm https://raw.githubusercontent.com/vinhnx/vtcode/main/scripts/install.ps1 | iex

Via Cargo:

cargo install vtcode

Via Homebrew:

brew install vtcode  # from homebrew/core
# or from vinhnx/tap (development):
brew tap vinhnx/tap
brew install vtcode

Usage

Set your API key and launch:

export OPENAI_API_KEY="sk-..."
vtcode

To install the optional search tools bundle (ripgrep + ast-grep) after initial installation:

vtcode dependencies install search-tools

Architecture

The project is organized into many Rust crates (e.g., vtcode-core, vtcode-llm, vtcode-tui, vtcode-tools) and includes a collaborative tool spec system, markdown store, config, and exec events. The agent loop is documented in the README and DeepWiki.

For development, see the GitHub repo for setup scripts (scripts/setup-ghostty-vt-dev.sh, etc.) and the Homebrew tap for bleeding-edge builds.

📖 Read the full source: HN AI Agents

Ad

👀 See Also

🦀
Tools

Claude Prototypes Real Estate Analysis App in 3 Hours Using Live Zillow Data via clawhub

A developer used Claude with the zillow-full clawhub tool to build a rental cash flow analysis app — pulling live Zillow API data, prototyping the UI around real JSON responses, and delivering a working prototype in one afternoon.

OpenClawRadar
DoomVLM: Open Source Tool for Testing Vision Language Models in Doom Deathmatches
Tools

DoomVLM: Open Source Tool for Testing Vision Language Models in Doom Deathmatches

DoomVLM is now open source as a single Jupyter notebook that lets you test vision language models playing Doom via OpenAI-compatible APIs. The tool supports deathmatch modes where up to 4 models can compete, with full configuration options for system prompts, tool descriptions, and sampling parameters.

OpenClawRadar
T9OS: An AI Orchestration System Built Entirely with Claude Code
Tools

T9OS: An AI Orchestration System Built Entirely with Claude Code

An economics student built T9OS, a complete AI orchestration layer using Claude Code as the only programming tool. The system includes 18 production pipelines, a 12-state lifecycle engine, and 7 AI 'Guardians' that review every output.

OpenClawRadar
Netflix Releases VOID: Video Object and Interaction Deletion Model on Hugging Face
Tools

Netflix Releases VOID: Video Object and Interaction Deletion Model on Hugging Face

Netflix has released VOID, a video inpainting model that removes objects from videos along with all physical interactions they induce, including falling objects and displaced items. The model requires a GPU with 40GB+ VRAM and uses quadmask conditioning with two checkpoint files for different refinement levels.

OpenClawRadar