Skip to main content
Settings
Search
Appearance
Theme Mode
About
Jekyll v3.10.0
Environment Production
Last Build
2026-05-22 22:41 UTC
Current Environment Production
Build Time May 22, 22:41
Jekyll v3.10.0
Build env (JEKYLL_ENV) production
Quick Links
Page Location
Page Info
Layout article
Collection posts
Path _posts/2026-05-17-mcp-servers-and-agent-tooling-in-practice.md
URL /posts/mcp-servers-and-agent-tooling-in-practice/
Date 2026-05-17
Theme Skin
SVG Backgrounds
Layer Opacity
0.6
0.04
0.08

MCP Servers and Agent Tooling in Practice

By IT-Journey Team

A practical guide to the Model Context Protocol and GitHub-native agent tooling — what MCP servers do and how GH-600 tests them.

Estimated reading time: 6 minutes

MCP Servers and Agent Tooling in Practice

Domain 2 of GH-600 (18% of the exam) covers the tools, environments, and capabilities that GitHub Copilot agents can use. The centrepiece is the Model Context Protocol (MCP).

What Is MCP?

MCP is an open standard that allows AI models to connect to external tools in a structured, predictable way. Think of it as a universal adapter. Without MCP, every tool integration requires custom code. With MCP, a tool exposes a standard interface, and any MCP-compatible model can use it.

The GitHub MCP server is one of the most important implementations. It gives agents the ability to:

  • Read and create issues
  • Read and update pull requests
  • Query repository contents
  • Check workflow run statuses
  • Manage labels and milestones

All without writing custom API code in every agent.

Configuring MCP in VS Code

MCP servers are configured in .vscode/mcp.json (workspace level) or VS Code user settings:

{
  "servers": {
    "github": {
      "command": "npx",
      "args": ["@modelcontextprotocol/server-github"],
      "env": {
        "GITHUB_PERSONAL_ACCESS_TOKEN": "${input:github-token}"
      }
    }
  }
}

The ${input:github-token} pattern prompts for the token securely, rather than hardcoding it.

Scoped vs. Unscoped Tools

Domain 2 sub-skill 2.1 covers tool selection and permissions. The principle: agents should only have access to tools they need for the current task. This is analogous to least-privilege in security.

For a code review agent:

  • ✅ Read repository contents
  • ✅ Create PR review comments
  • ❌ Create issues (not needed)
  • ❌ Manage repo settings (not needed)

Tool scoping reduces the blast radius of an agent that misbehaves or is given ambiguous instructions.

Development Environment Integration (Sub-skill 2.3)

Sub-skill 2.3 covers how agents interact with the development environment. This includes:

  • AGENTS.md — a file the agent reads to understand the repository conventions, forbidden actions, and preferred patterns
  • .github/copilot-instructions.md — instructions that Copilot reads to understand the project context
  • Git configuration — ensuring the agent uses a clearly identified committer identity

Domain 2 Quests

Quest Skill Link
Q4 Tool Selection & Permissions Tool Selection & Permissions
Q5 MCP Server Mastery MCP Server Mastery
Q6 Dev Environment Integration Dev Environment Integration
Q7 Safe Execution & Error Handling Safe Execution & Error Handling

These quests cover MCP configuration, token scoping, AGENTS.md authoring, and error escalation workflows.