GitHub Copilot SDK Configuration

Configure thresh to work with GitHub Copilot Chat in VS Code, enabling AI-powered environment management through natural language commands.

What You'll Learn

• Configure thresh as an MCP server for GitHub Copilot
• Use natural language to manage environments
• Generate blueprints with AI assistance
• Troubleshoot AI chat sessions
• Best practices for AI-driven development

Prerequisites

• thresh installed (Quick Start)
• VS Code with GitHub Copilot extension
• Active GitHub Copilot subscription
• Basic familiarity with VS Code

Overview

thresh implements the Model Context Protocol (MCP), allowing GitHub Copilot to:

• List and provision environments
• Generate custom blueprints
• Manage environment lifecycle
• Answer questions about thresh capabilities

You (VS Code) → GitHub Copilot → thresh MCP Server → Container Runtime

Step 1: Initialize MCP Configuration

Run the automatic configuration tool:

thresh index

Output:

[INFO] Initializing MCP configuration for VS Code

Detected workspace: C:\Users\user\projects\my-app

Creating MCP configuration:
  File: .vscode\mcp.json
  Server: thresh
  Command: thresh serve
  
✓ MCP configuration created successfully!

GitHub Copilot will now have access to thresh commands.
Reload VS Code or GitHub Copilot to activate.

This creates .vscode/mcp.json:

{
  "mcpServers": {
    "thresh": {
      "command": "thresh",
      "args": ["serve"]
    }
  }
}

info

thresh index automatically detects your VS Code workspace and creates the configuration. You can run it in any project directory.

Step 2: Reload VS Code

For the MCP configuration to take effect:

1. Command Palette: Press Ctrl+Shift+P
2. Reload Window: Type "Reload Window" and press Enter

Alternatively:

• Restart GitHub Copilot: Ctrl+Shift+P → "GitHub Copilot: Reload Copilot"

Step 3: Verify Connection

Open GitHub Copilot Chat:

• Shortcut: Ctrl+Alt+I
• Or: Click the chat icon in the sidebar

Test the connection:

Can you list my thresh environments?

Expected response:

I can see 2 thresh environments:

1. python-dev (running)
   - Alpine 3.19
   - Uptime: 2h 15m

2. node-dev (stopped)
   - Alpine 3.19
   - Last active: 1 day ago

Would you like me to start the node-dev environment?

Using AI Commands

List Environments

Show me all my thresh environments

What environments are running?

Provision Environment

Create a Python development environment

I need a Node.js environment with TypeScript

Generate Custom Blueprint

Generate a blueprint for Django development with PostgreSQL and Redis

Copilot will:

1. Analyze your requirements
2. Choose appropriate distribution
3. Select packages
4. Configure environment variables
5. Create blueprint JSON
6. Save to .thresh/blueprints/

Get Environment Information

What's the status of my python-dev environment?

How much memory is node-dev using?

Troubleshooting

Why won't my python-dev environment start?

How do I install additional packages in my environment?

Real-World Examples

Example 1: Create Backend API

You:

I'm building a FastAPI backend. Create an environment with Python 3.11, 
FastAPI, uvicorn, and PostgreSQL client tools.

Copilot:

I'll create a custom environment for FastAPI development.

Creating blueprint: fastapi-backend.json...

{
  "distribution": "alpine:3.19",
  "packages": ["python3", "py3-pip", "postgresql-client"],
  "postInstall": [
    "pip install fastapi uvicorn sqlalchemy psycopg2-binary"
  ]
}

Provisioning environment...
✓ fastapi-backend environment ready!

You can enter it with: wsl -d thresh-fastapi-backend

Example 2: Multi-Environment Setup

You:

Set up environments for a microservices project:
- Frontend: Node.js 20 with React
- Backend: Python with Flask
- Database: Just the PostgreSQL client tools

Copilot:

I'll create 3 environments for your microservices architecture:

1. Creating react-frontend...
   - Node.js 20, npm, create-react-app
   
2. Creating flask-backend...
   - Python 3.11, Flask, SQLAlchemy
   
3. Creating db-tools...
   - PostgreSQL client
   
All environments provisioned successfully!

Use:
  wsl -d thresh-react-frontend
  wsl -d thresh-flask-backend
  wsl -d thresh-db-tools

Example 3: Debug Issues

You:

My python-dev environment won't start. What's wrong?

Copilot:

Let me check the environment status...

Issue found: WSL 2 distribution is corrupted

Fix:
1. Export any important files (they'll be in /home/user)
2. Destroy the environment: thresh destroy python-dev
3. Recreate it: thresh up python-dev

Would you like me to help you backup files first?

Advanced Configuration

Custom Server Path

If thresh isn't in PATH, specify full path:

.vscode/mcp.json:

{
  "mcpServers": {
    "thresh": {
      "command": "C:\\Program Files\\thresh\\thresh.exe",
      "args": ["serve"]
    }
  }
}

Environment Variables

Pass configuration to thresh server:

{
  "mcpServers": {
    "thresh": {
      "command": "thresh",
      "args": ["serve"],
      "env": {
        "THRESH_CONFIG": "C:\\custom\\config.json",
        "THRESH_LOG_LEVEL": "debug"
      }
    }
  }
}

Logging

Enable debug logs for troubleshooting:

{
  "mcpServers": {
    "thresh": {
      "command": "thresh",
      "args": ["serve", "--log-file", "C:\\thresh-mcp.log", "--log-level", "debug"]
    }
  }
}

View logs:

Get-Content C:\thresh-mcp.log -Wait

Multi-Workspace Setup

Different projects can have different MCP configurations:

Projects/
├── web-app/
│   └── .vscode/
│       └── mcp.json  ← thresh for this project
├── mobile-app/
│   └── .vscode/
│       └── mcp.json  ← thresh + other tools
└── data-science/
    └── .vscode/
        └── mcp.json  ← thresh with custom blueprints

Workspace vs User Settings

Workspace (.vscode/mcp.json)

Best for: Project-specific configuration, team sharing

{
  "mcpServers": {
    "thresh": {
      "command": "thresh",
      "args": ["serve"]
    }
  }
}

Committed to Git for team use.

User Settings

Best for: Personal, global configuration across all projects

File: %APPDATA%\Code\User\settings.json (Windows)

{
  "github.copilot.advanced": {
    "mcp.thresh.command": "thresh",
    "mcp.thresh.args": ["serve"]
  }
}

AI Chat Best Practices

Be Specific

❌ Vague:

Make an environment

✅ Specific:

Create a Python 3.11 environment with Flask, SQLAlchemy, and pytest

Provide Context

❌ No context:

It's not working

✅ With context:

My python-dev environment won't start. 
Error: "WSL 2 kernel not found"
OS: Windows 11

Use Natural Language

You don't need exact command syntax:

✅ "Show my environments"
✅ "List all environments"
✅ "What environments do I have?"
✅ "Are any environments running?"

All produce the same result!

Iterate

You: Create a Node.js environment
Copilot: [creates basic Node environment]

You: Add TypeScript and ESLint
Copilot: [updates blueprint with TypeScript]

You: Also add Jest for testing
Copilot: [adds Jest configuration]

Troubleshooting

Copilot Doesn't See thresh

Symptom: Copilot says "I don't have access to thresh commands"

Fix:

1. Verify .vscode/mcp.json exists in workspace
2. Reload VS Code window
3. Check thresh is in PATH: thresh --version
4. Restart GitHub Copilot extension

MCP Server Won't Start

Symptom: "Failed to start MCP server"

Debug:

# Test server manually
thresh serve

# Should show:
# [INFO] MCP server starting on stdio transport
# [INFO] Server ready, waiting for requests...

If this fails, check:

• thresh installation: thresh --version
• Container runtime: wsl --status

Slow Responses

Symptom: Copilot takes more than 10 seconds to respond

Causes:

• First request downloads distribution images
• Large number of environments (more than 10)
• Container runtime initialization

Fix:

• Subsequent requests are faster (caching)
• Destroy unused environments: thresh list → thresh destroy <name>

Permission Errors

Symptom: "Access denied" or "Permission denied"

Windows WSL:

# Ensure WSL 2 is running
wsl --status

# Restart WSL if needed
wsl --shutdown
wsl

Examples Library

Data Science

Create a Jupyter notebook environment with pandas, NumPy, 
matplotlib, and scikit-learn

Web Development

I need a full-stack environment: Node.js frontend + Python backend

DevOps

Set up an environment with Docker, kubectl, Terraform, and Azure CLI

Mobile Development

Create a React Native development environment with Android SDK tools

Comparison: CLI vs AI Chat

Task	CLI Command	AI Chat
List environments	thresh list	"Show my environments"
Create environment	thresh up python-dev	"Create a Python environment"
Custom blueprint	Edit JSON manually	"Generate a Django blueprint"
Troubleshoot	Read logs, docs	"Why won't my env start?"
Learn	Read CLI reference	"How do I add packages?"

Use CLI when: Speed, scripting, automation
Use AI when: Learning, complex setups, troubleshooting

Security Considerations

What Copilot Can Access

Copilot (via thresh MCP) can:

• ✅ List environments
• ✅ Create/destroy environments
• ✅ Read blueprint files
• ❌ Access files inside environments
• ❌ Modify container runtime directly

Data Privacy

• Environment names and metadata sent to GitHub Copilot API
• Blueprint templates may be analyzed by AI
• No source code from environments is sent
• MCP communication is via local stdio (no network)

Disable MCP

Remove .vscode/mcp.json to disable thresh integration.

Next Steps

• VS Code MCP Integration - Deep dive into MCP protocol
• Creating Custom Blueprints - Manual blueprint creation
• CLI Reference: serve - MCP server details

Community Examples

Share your Copilot + thresh workflows on GitHub Discussions!

---

You can now use natural language to manage development environments through GitHub Copilot. Experiment with different prompts to discover what's possible!