🤖 GitHub Copilot SDK / 90-minute core workshop

Build an AI-powered accessibility reviewer

The GitHub Copilot SDK lets your application run Copilot as an agent: send prompts, stream responses, expose application functions as tools, and connect external capabilities through MCP.

You'll build a console application that accepts a webpage URL, opens it with Playwright, identifies accessibility problems, consults an application-owned Web Content Accessibility Guidelines (WCAG) catalog, and produces a structured report.

URL
Playwright inspection
WCAG lookup
Accessibility report

🚀 Start workshop

Choose a language first. No agent, tool-calling, or MCP experience required.

accessibility-report

URL target page

[tool] playwright-browser_navigate
[tool] read_latest_accessibility_snapshot
[tool] accessibility_rule_lookup

Finding
The name input has no accessible name.

WCAG
4.1.2 Name, Role, Value

Fix
Associate a visible <label> with the input.

By the end

Build the app and understand the boundaries behind it.

Create and manage a Copilot session.
Stream responses through session events.
Expose application-owned knowledge as a local tool.
Connect Playwright MCP through scoped permissions.
Separate browser evidence from model interpretation.

Build an AI-powered accessibility reviewer

Choose your workshop language

Build the app and understand the boundaries behind it.