GitHub Copilot MCP Integration

GitHub Copilot MCP Integration

Prompt-first procedure: Describe the outcome you want in your agent conversation. The agent should select and load the appropriate AIWG assets, explain material changes, request any needed approval, and report verification evidence. Exact commands and flags appear only in the CLI reference.

Status: Supported — MCP is GA in both VS Code agent mode and the Copilot coding agent.

Overview

GitHub Copilot supports MCP (Model Context Protocol) in two distinct contexts:

ContextConfigurationCapabilities
VS Code (Agent Mode + Chat)`.vscode/mcp.json`Tools, prompts, resources
Coding Agent (GitHub.com)GitHub.com Settings UITools only

VS Code MCP Configuration

Setup

Add AIWG as an MCP server in `.vscode/mcp.json`:

{
  "servers": {
    "aiwg": {
      "type": "stdio",
      "command": "aiwg",
      "args": ["mcp", "serve"]
    }
  }
}

Server Types

TypeUse CaseExample
`stdio`Local process (recommended for AIWG)`"command": "aiwg", "args": ["mcp", "serve"]`
`http`Remote server with streamable HTTP`"url": "https://mcp.example.com/mcp"`
`sse`Server-Sent Events (legacy)`"url": "https://mcp.example.com/sse"`

Authentication

For servers requiring authentication, use input variables:

{
  "servers": {
    "authenticated-server": {
      "type": "http",
      "url": "https://api.example.com/mcp",
      "headers": {
        "Authorization": "Bearer ${input:api-key}"
      }
    }
  },
  "inputs": [
    {
      "type": "promptString",
      "id": "api-key",
      "description": "Your API Key",
      "password": true
    }
  ]
}

What MCP Enables

Once configured, Copilot agent mode can:

  • Call AIWG tools (`discover`, `command-run`, `artifact-read`, etc.)
  • Use MCP prompts (invokable as `/mcp.aiwg.promptname`)
  • Access MCP resources for context
  • Chain multiple MCP servers together

Tools are auto-invoked based on prompt intent. Non-read-only tools show confirmation dialogs.

User-Level Configuration

For personal MCP servers across all projects, use the VS Code command: `MCP: Open User Configuration`

Agent-Scoped MCP

Custom agents (`.github/agents/*.agent.md`) can declare their own MCP servers:

---
name: data-analyst
mcp-servers:
  - type: stdio
    command: npx
    args: ['-y', '@company/data-mcp']
---

Sandbox

On macOS/Linux, MCP servers run with configurable filesystem and network access rules for security isolation.

Coding Agent MCP Configuration

Setup

Configure via GitHub.com > Settings > Copilot > Coding agent:

{
  "mcpServers": {
    "sentry": {
      "type": "http",
      "url": "https://mcp.sentry.io",
      "tools": ["get_issue_details", "get_issue_summary"]
    },
    "local-tool": {
      "type": "local",
      "command": "npx",
      "args": ["-y", "@company/tool"],
      "tools": "*"
    }
  }
}

Key Differences from VS Code

AspectVS Code MCPCoding Agent MCP
Configuration`.vscode/mcp.json` (committed)GitHub.com Settings UI
CapabilitiesTools + prompts + resourcesTools only
ApprovalUser confirms non-read-only toolsAutonomous (no approval)
ScopeWorkspace or user-levelRepository or org-level

Required Fields

  • `type`: `http` or `local`
  • `tools`: Array of tool names or `"*"` for all tools from server

AIWG MCP Server

AIWG provides a built-in MCP server exposing SDLC workflow tools:

Use AIWG to complete this documented outcome: AIWG provides a built-in MCP server exposing SDLC workflow tools
Have it inspect the current state, explain the plan, ask before material
changes, and report the result with verification evidence.

Exposed Tools

The AIWG MCP server exposes tools for:

  • Discovery and CLI dispatch (`discover`, `command-run`)
  • Artifact management (`artifact-read`, `artifact-write`)
  • Catalog lookup (`skill-`, `command-`, `rule-`, `agent-`, `template-*`)
  • Optional Flow and Mission tools when `AIWG_MCP_TOOLSETS=flows,missions`

For Teams Using VS Code + Copilot

1. Deploy AIWG agents and prompts:

Use AIWG to complete this documented outcome: 1. Deploy AIWG agents and prompts
Have it inspect the current state, explain the plan, ask before material
changes, and report the result with verification evidence.

2. Add MCP configuration:

Use AIWG to complete this documented outcome: 2. Add MCP configuration
Have it inspect the current state, explain the plan, ask before material
changes, and report the result with verification evidence.

3. Commit both `.github/` and `.vscode/mcp.json`

4. Team members get full AIWG integration:

  • Custom agents via @-mention
  • Custom slash commands via /name
  • MCP tools via agent mode

For Teams Using Copilot Coding Agent

1. Deploy AIWG agents:

Use AIWG to complete this documented outcome: 1. Deploy AIWG agents
Have it inspect the current state, explain the plan, ask before material
changes, and report the result with verification evidence.

2. Configure MCP in GitHub.com Settings (if using external tools)

3. The coding agent reads `.github/copilot-instructions.md` for project context