Contents · What is Fetch Coder? · Quick start · Features · Configuration · Keyboard shortcuts · Toolbar guide · Install from source · Documentation · Resources · Store listings and releases · License

🎯 What is Fetch Coder?

Fetch Coder is an autonomous AI coding agent for VS Code and Cursor, powered by ASI:One (OpenAI-compatible API at https://api.asi1.ai/v1). It plans and executes changes in your repo: read files, run terminal commands, and iterate until the task is done.

Perfect for:

• 🏗️ Building full-stack applications from scratch
• 🐛 Debugging complex issues across your codebase
• 📝 Refactoring and optimizing code
• 🔧 DevOps tasks (Docker, Kubernetes, CI/CD)
• 🤖 Creating autonomous agents and AI systems
• 📚 Learning new frameworks and technologies

⚡ Quick Start

1️⃣ Install

Search "Fetch Coder" in the Extensions view (Cursor or VS Code), or install from VS Code Marketplace.

2️⃣ Get API Access

1. Open the ASI:One developer portal (create and manage API keys).
2. Optionally follow the API quickstart.
3. Paste the key in Fetch Coder’s welcome screen or Settings → API Configuration.

3️⃣ Start Building

Open the Fetch Coder sidebar and describe what you need:

"Create a modern landing page with animations and dark mode"
"Build a REST API with authentication and database"
"Debug my React app's performance issues"
"Write tests for my Python module"
"Set up Docker, GitHub Actions, and deployment"

✨ Features

💬 Intelligent Chat

Chat with asi1.ai naturally. It understands your code, explains problems, suggests improvements, and executes changes—all in real-time with streaming responses.

🎨 Beautiful Welcome Screen

• asi1.ai logo & branding prominently displayed
• Gradient text headers with modern design
• Quick-start suggested tasks with icons
• Get started immediately with pre-built examples
• Settings quick access for prompt customization

📝 Smart Code Generation

• Write new features
• Fix bugs automatically
• Refactor existing code
• Generate tests and documentation
• Create full applications

🗂️ File Management

• Read and analyze files
• Create new files and folders
• Modify multiple files at once
• See diffs before applying changes
• Auto-rollback if something breaks

💻 Terminal Integration

Run any terminal command and read output:

• npm install, git commit, docker build
• Python, Node.js, bash scripts
• Database migrations
• Long-running processes with streaming output

🔍 Web Search

Toggle the globe icon to enable real-time web search. Fetch Coder can research current best practices, APIs, and solutions while coding.

✨ Prompt Enhancement

Click the sparkles icon to let ASI:One rewrite your prompt for better results with more details and clarity.

📊 Plan & Act Modes

• Plan Mode: Analyze the problem, gather info, design a solution
• Act Mode: Execute changes immediately

Toggle with the Plan/Act switch in the chat input.

🔌 MCP Server Support

Connect to specialized tools and services via Model Context Protocol:

• Databases (PostgreSQL, MongoDB, etc.)
• APIs and integrations
• Custom tools and services
• Real-time data access

💾 Checkpoints & History

• Auto-save checkpoints before changes
• Revert to any previous state
• Full task history with search
• Resume interrupted conversations

🤖 Subagents

For complex tasks, spawn parallel subagents to work on different parts simultaneously—faster development.

🌐 Browser Automation

Automated testing and interaction:

• Launch headless browser
• Navigate and click elements
• Fill forms, submit data
• Verify UI changes

🛠️ Configuration

Branding (optional Cursor-style accents)

• Settings → search fetchCoder.brandTheme: cursor-accent (default) uses blue/cyan marketing accents; host uses the classic ASI blue/lime. Editor colors still follow your theme via VS Code tokens.
• Optional: fetchCoder.hostApp: auto (use vscode.env.appName), vscode, or cursor to fix wrong host detection for placeholders and welcome copy. This does not integrate with the editor’s native chat.

Extra system instructions

• Settings → fetchCoder.extraSystemPrompt: optional text appended to the built-in system prompt on every model request (max length enforced in the extension).

API Key Setup (Choose one)

Option 1: Welcome Screen (Recommended)

• Paste when Fetch Coder first opens

Option 2: Command Palette

Cmd+Shift+P → "Fetch Coder: Set ASI:One API Key & Base URL" (or "ASI: Set API Key")

Option 3: VS Code Settings (synced to ASI:One)

{
  "asiAssistant.apiKey": "your-api-key-here",
  "asiAssistant.baseUrl": "https://api.asi1.ai/v1",
  "asiAssistant.model": "asi1"
}

Option 4: Environment variable

export ASI_ONE_API_KEY="your-api-key-here"

Model & endpoint (ASI:One)

• Base URL (one OpenAI-compatible endpoint for chat): https://api.asi1.ai/v1 — override with asiAssistant.baseUrl if your deployment differs.
• Model (dropdown in the extension, synced from asiAssistant.model): asi1-ultra, asi1, or asi1-mini (default: asi1).

Privacy & security

• Your project files are processed locally; API requests for completions are sent to the ASI:One endpoint you configure.
• Optional telemetry — In Fetch Coder → Settings → General, Allow error and usage reporting toggles aggregated usage and diagnostics. When unchecked, reporting is suppressed. The in-extension description lists what may be collected; code, prompts, and personal information are not sent in that stream.
• Editor telemetry — VS Code’s Telemetry: Telemetry Level (or Cursor’s equivalent) can also disable client-side reporting that extensions depend on.
• For extension log output, use View → Output and choose the Asi channel in the dropdown.

⌨️ Keyboard Shortcuts

Use View → Open View… → Fetch Coder or the Activity Bar icon to open the sidebar. Plan/Act is toggled in the chat UI (no default keybinding for it).

Not on the roadmap (yet)

The extension does not sync threads with Cursor’s built-in chat, hook Cursor’s inline / commands, or replace the host’s gRPC streaming stack with a raw webview fetch client — those would need official APIs and/or a larger product decision. The sidebar agent remains the integration surface.

🎨 Toolbar Guide

📦 Installation from Source

# Clone repository
git clone https://github.com/gautammanak1/asi1-vs-code.git
cd asi1-vs-code

# Install dependencies (root + webview-ui)
npm run install:all

# Build
npm run protos
npm run build:webview
npm run compile

# Launch (press F5)
# Opens VS Code Extension Development Host

📚 Documentation

Project guides (beyond this README):

🔗 Resources

Store listings and releases

VS Code Marketplace: The published extension’s long description is this root README. The storefront name, short summary, and version come from package.json (displayName, description, version).

Open VSX: If you publish the same .vsix to Open VSX, the same README and package.json fields apply once the registry has ingested your upload.

Cursor: Cursor installs standard VS Code extensions (View → Extensions). Listing text reflects the package you publish. This extension’s chat runs in the sidebar webview; it does not integrate with Cursor’s built-in chat—see Not on the roadmap (yet).

Release checklist (maintainers)

• Bump version in package.json and refresh README release badges and the “Current extension version” subtitle if they pin a specific SemVer string.
• Keep description in package.json aligned with the one-line summary shown in marketplace search results (this README is the long description below it).
• Add an entry in CHANGELOG.md for the release.
• Run your usual npm build and publish flow (VS Code Marketplace publisher and Open VSX, if applicable). After publishing, verify the storefronts show the updated short summary and README.

📄 License

Apache License 2.0 — see LICENSE for details.