Copilot mcp server

Copilot MCP Server

When to use this skill:

• When you need direct access to Copilot AI models for specific tasks
• When you want to use Copilot's specialized tools (review, debug, explain, etc.)
• When you need AI assistance outside of the copilot-flow workflow

Available MCP Tools

Core Tools

ask-copilot

General-purpose AI assistant for coding help, debugging, and architecture design.

mcp__plugin__copilot__ask-copilot(
  prompt="string",           // Required: The question or task for Copilot
  context="string",          // Optional: Additional context
  model="string",            // Optional: Specific model to use (default: claude-sonnet-4.6)
  allowAllTools=true/false   // Optional: Allow Copilot to use all available tools
)

copilot-review

Professional code review with focus on specific areas.

mcp__plugin__copilot__copilot-review(
  code="string",                      // Required: Code to review
  focusAreas=["security", "performance", "maintainability", "best-practices"]  // Optional: Specific areas to focus
)

copilot-explain

Get detailed explanations of code snippets.

mcp__plugin__copilot__copilot-explain(
  code="string",      // Required: Code to explain
  model="string"      // Optional: Model to use
)

copilot-debug

Debug errors in code with context-aware analysis.

mcp__plugin__copilot__copilot-debug(
  code="string",      // Required: Code with error
  error="string",     // Required: Error message
  context="string"    // Optional: Additional context
)

copilot-refactor

Get suggestions for code refactoring and improvements.

mcp__plugin__copilot__copilot-refactor(
  code="string",      // Required: Code to refactor
  goal="string"       // Required: Refactoring goal (e.g., "improve performance")
)

copilot-test-generate

Generate unit tests for existing code.

mcp__plugin__copilot__copilot-test-generate(
  code="string",        // Required: Code to test
  framework="string"    // Optional: Testing framework (e.g., jest, pytest, mocha)
)

copilot-suggest

Get CLI command suggestions for specific tasks.

mcp__plugin__copilot__copilot-suggest(
  task="string",       // Required: Task description
  model="string"       // Optional: Model to use
)

Session Management

copilot-session-start

Start a new conversation session with context tracking.

mcp__plugin__copilot__copilot-session_start()

copilot-session-history

Retrieve conversation history for continuity.

mcp__plugin__copilot__copilot-session_history(
  sessionId="string"   // Optional: Specific session ID
)

Model Selection

Choose from available models based on task complexity:

claude-sonnet-4.6 (default)

• Best for: System design, architecture decisions, code review, performance optimization
• Balance of capability and speed

claude-opus-4.7

• Best for: Complex problems requiring highest reasoning on eligible plans
• Optional higher-tier model, not the default

claude-haiku-4.5

• Best for: Quick syntax queries, simple logic questions, API usage
• Fastest response

gpt-5.4

• Best for: Broad code generation, debugging, and architecture tradeoff analysis
• Strong general-purpose GPT option

gpt-5.3-codex

• Best for: Codex-style code generation workflows available in the current CLI picker
• Useful when you want a GPT coding-focused alternative to GPT-5.4

gpt-5-mini (unlimited usage)

• Best for: Concept explanations, general technical questions, documentation queries
• No usage limits for Pro+ subscribers

gpt-5.2-codex

• Best for: Complex algorithms, system refactoring, large feature development
• Advanced code generation

gpt-5.2

• Best for: Complex reasoning with large context
• Highest comprehension

Usage Examples

Code Implementation

mcp__plugin__copilot__ask-copilot(
  prompt="Implement a REST API endpoint for user authentication with JWT",
  model="claude-sonnet-4.6",
  allowAllTools=true
)

Security Review

mcp__plugin__copilot__copilot_review(
  code=`function login(username, password) {
    const query = \`SELECT * FROM users WHERE username = '\${username}' AND password = '\${password}'\`;
    return db.query(query);
  }`,
  focusAreas=["security", "sql-injection", "authentication"]
)

Error Debugging

mcp__plugin__copilot__copilot_debug(
  code="const result = await fetchData().json;",
  error="TypeError: fetchData(...).json is not a function",
  context="Trying to parse JSON response from API"
)

Test Generation

mcp__plugin__copilot__copilot_test_generate(
  code=`function isPrime(n) {
    if (n <= 1) return false;
    for (let i = 2; i * i <= n; i++) {
      if (n % i === 0) return false;
    }
    return true;
  }`,
  framework="jest"
)

Best Practices

Do

• Be specific in your prompts
• Provide context when available
• Choose appropriate model for task complexity
• Use session management for related queries
• Focus review on specific areas

Don't

• Use overly broad prompts like "write code"
• Overuse powerful models for simple tasks
• Pass empty parameters
• Ask for review on all areas without focus
• Mix unrelated queries in one session

Error Handling

If MCP server is unavailable:

1. Check GitHub Copilot CLI authentication
2. Verify MCP server configuration
3. Restart Claude Code or your MCP client

Integration Notes

• This skill provides direct tool access
• For structured workflow, use copilot-flow-integration skill
• All tools require active GitHub Copilot subscription
• Usage counts towards Copilot API limits

Keywords

copilot, mcp, ai, code review, debugging, testing, refactoring, github copilot, claude, gpt