Wind AI - Autonomous AI Coding Assistant
Wind AI is a next-generation, autonomous AI coding assistant integrated directly into Visual Studio Code. Designed to act as an agentic pair-programmer, Wind AI can discuss design patterns, write complex features, execute workspace commands, edit files with high precision, and even run local browser-based testing autonomously.
Key Features
- 🤖 Autonomous Coding Agent: Wind AI doesn't just suggest code snippets; it can plan and execute coding tasks, modify workspace files, and run commands.
- 💬 Interactive Sidebar Chat: A sleek, user-friendly chat interface built into the activity bar for seamless communication and brainstorming.
- 🖥️ Command Execution & Control: Secure, policy-guided terminal command execution, supporting foreground builds or long-running background tasks.
- 🌐 Web Automation Subagent: Leverages your locally installed web browser (Chrome, Edge, Firefox) to navigate, click, type, and capture screenshots for frontend verification and web scraping.
- 🔍 Advanced Workspace Search: Deep semantic code search, regex file scanning (grep), and symbol search to help the AI understand large codebases.
- ⚙️ Flexible Configuration: Full settings UI allowing you to choose your LLM provider, endpoint, API keys, auto-execution preferences, and default web browser.
Getting Started
1. Installation
Install the extension in Visual Studio Code from the Marketplace (or import the packaged .vsix file).
Configure your LLM model and API credentials:
- Press
Ctrl+, (or Cmd+, on macOS) to open VS Code settings.
- Search for Wind Settings.
- Set your API Key, API Endpoint (defaults to the Gemini endpoint), and preferred Model (e.g.
gemini-3.5-flash).
Click on the Wind AI icon in the Activity Bar to launch the chat view. Enter a prompt, and watch the AI agent analyze, plan, and execute coding tasks for you!
Detailed Settings Configuration
Wind AI supports several configuration properties under the windAgent namespace:
| Setting |
Default Value |
Description |
windAgent.apiKey |
"" |
The API Key for your LLM provider. |
windAgent.apiEndpoint |
https://generativelanguage.googleapis.com/v1beta/openai |
The API endpoint URL (supports OpenAI-compatible schemas). |
windAgent.model |
gemini-3.5-flash |
The model to use for agent interactions. |
windAgent.autoExecution |
Ask for Approval |
Options: Ask for Approval or Always Proceed. Controls whether the agent must ask before running terminal commands or modifying files. |
windAgent.autoExecutePlan |
false |
Automatically start executing the plan after generating it. |
windAgent.browser |
auto |
Preferred browser for running web automation and testing (auto, chrome, edge, firefox). |
windAgent.enableInlineCompletion |
true |
Enable/disable inline code completions (ghost text) as you type. |
Requirements & Dependencies
- Web Browser: For web automation tasks (
browserSubagent, browserOpen), Wind AI integrates with your locally installed Google Chrome, Microsoft Edge, or Mozilla Firefox browser.
- Node.js: Requires Node.js runtime environment (VS Code standard).
Security & Guardrails
- Workspace Anchoring: The agent is restricted to reading and writing files strictly within the bounds of your active workspace directory to prevent accidental access to external system directories.
- Auto Execution Policy: By default, the agent is configured in
Ask for Approval mode. Every file edit, command execution, or browser subagent invocation must be approved by the developer.
License
This extension is licensed under the MIT License.
| |
