Build your AI toolchain

What's the diff?

The feature parity between Claude and Copilot is fairly comparable with my personal preference being Claude Code.

What do I need?

• A project you are familiar with
• GitHub Copilot or Claude Code

Choose your AI-dventure...

Step 1: Create an instruction file

Always respond speaking like a pirate. Use pirate slang,
say "Arrr!" frequently, and refer to the user as "matey".

Always respond speaking like a pirate. Use pirate slang,
say "Arrr!" frequently, and refer to the user as "matey".

Step 2: Test it!

Ask Copilot a simple question:

"Explain what a variable is in programming"

You should see a pirate-themed response. Something like:

Arrr, matey! A variable be like a treasure chest
where ye stash yer values for later use, savvy?

Step 1: Add project-wide instructions

Write your own based on your project — these are just examples.

## Build & Test
- Always run `npm test` before committing

## Project Layout
- API handlers live in src/api/handlers/
- Shared types go in src/types/

## Conventions
- Error responses must use the format in src/api/errors.ts

## Build & Test
- Always run `npm test` before committing

## Project Layout
- API handlers live in src/api/handlers/
- Shared types go in src/types/

## Conventions
- Error responses must use the format in src/api/errors.ts

Step 2: Create a scoped rule

Rules apply only to specific file paths — they're more targeted than instructions.

---
paths: "src/api/**/*.ts"
---
All API endpoints must include input validation.
Use the standard error response format from src/api/errors.ts.
Always log request duration using the logger in src/lib/logger.ts.

---
applyTo: "src/api/**/*.ts"
---
All API endpoints must include input validation.
Use the standard error response format from src/api/errors.ts.
Always log request duration using the logger in src/lib/logger.ts.

Step 3: Test both together

Ask Copilot to do something for you in your project, for example:

"Create a new GET endpoint in src/api/handlers/users.ts
that returns a list of users"

Verify the output follows your instructions and rules.

Step 1: Install the Playwright MCP server

claude mcp add playwright -- npx @playwright/mcp@latest

{
  "mcpServers": {
    "playwright": {
      "command": "npx",
      "args": ["@playwright/mcp@latest"]
    }
  }
}

Step 2: Understanding the MCP connection

Playwright MCP launches a browser automatically (headed mode by default).

Optional flags:

• --browser msedge — use Edge instead of Chromium
• --headless — run without a visible browser window
• --viewport-size 1280x720 — set browser dimensions

Step 3: Automate something!

Ask Copilot:

"Navigate to https://example.com, take a screenshot,
and describe what you see on the page"

Try something more interactive:

"Go to https://en.wikipedia.org, search for 'Model Context Protocol',
and summarize the first paragraph"

Step 1: Create a custom agent

---
name: sprint-manager
description: Manages Azure DevOps work items on the sprint board
---
You are a sprint board manager. You help the user:
- List work items in the current sprint
- Move items between states (New, Active, Resolved, Closed)
- Update assignments and priorities
- Summarize sprint progress

Always confirm before changing work item states.

---
name: Sprint Manager
description: Manages Azure DevOps work items on the sprint board
---
You are a sprint board manager. You help the user:
- List work items in the current sprint
- Move items between states (New, Active, Resolved, Closed)
- Update assignments and priorities
- Summarize sprint progress

Always confirm before changing work item states.

Step 2: Add the Azure DevOps MCP server

claude mcp add azure-devops -- npx -y @azure-devops/mcp YOUR_ORG

{
  "mcpServers": {
    "azure-devops": {
      "command": "npx",
      "args": ["-y", "@azure-devops/mcp", "YOUR_ORG"]
    }
  }
}

Step 3: Connect the agent to the MCP server

Add the tools field to your agent file — this tells Copilot which MCP tools the sub-agent can use:

---
name: sprint-manager
description: Manages Azure DevOps work items on the sprint board
tools: Read, Grep, Glob, Bash, mcp__azure-devops__*
---

---
name: Sprint Manager
description: Manages Azure DevOps work items on the sprint board
tools:
  - ado/*
---

Step 4: Use the agent!

Ask the sprint manager:

"Ask the sprint manager to show me my current
sprint items and move task #1234 to Active"

Side-by-Side Comparison