MCP & Copilot Integration¶

Klondike integrates with GitHub Copilot CLI and provides an MCP (Model Context Protocol) server for AI agent tools.

GitHub Copilot Integration¶

Launch Copilot with Klondike Context¶

Start a Copilot session with full project context:

klondike copilot start

This launches Copilot CLI with:

Project status
Feature registry
Recent session history
Klondike instructions

Copilot Options¶

Option	Description	Example
`--model`	AI model to use	`claude-sonnet`, `gpt-4`
`--resume`	Resume previous session	Flag
`--feature`	Focus on specific feature	`F001`
`--instructions`	Additional instructions	`"Focus on tests"`
`--allow-tools`	Limit available tools	`"read,write"`
`--dry-run`	Show command without executing	Flag

Examples¶

# Start with specific model
klondike copilot start --model claude-sonnet

# Focus on a specific feature
klondike copilot start --feature F001

# Resume previous session
klondike copilot start --resume

# Preview command without running
klondike copilot start --dry-run

MCP Server¶

Klondike provides an MCP server that exposes project tools to AI agents.

Start the MCP Server¶

# Default: stdio transport
klondike mcp serve

# HTTP transport
klondike mcp serve --transport streamable-http

Install for VS Code¶

Configure VS Code to use the Klondike MCP server:

klondike mcp install

This updates your VS Code settings with the MCP server configuration.

Generate Config¶

Generate a configuration file for manual setup:

klondike mcp config --output mcp-config.json

Transport Options¶

Transport	Use Case	Command
`stdio`	Local development, CLI	`klondike mcp serve`
`streamable-http`	Remote access, web-based tools	`klondike mcp serve --transport streamable-http`

MCP Tools¶

The Klondike MCP server exposes these tools to AI agents:

Project Tools¶

Tool	Description
`klondike_status`	Get project status and feature summary
`klondike_validate`	Validate artifact integrity

Feature Tools¶

Tool	Description
`klondike_feature_list`	List all features
`klondike_feature_add`	Add a new feature
`klondike_feature_start`	Start working on a feature
`klondike_feature_verify`	Verify a completed feature
`klondike_feature_block`	Block a feature
`klondike_feature_show`	Show feature details

Session Tools¶

Tool	Description
`klondike_session_start`	Start a new session
`klondike_session_end`	End the current session

Progress Tools¶

Tool	Description
`klondike_progress`	Regenerate progress file
`klondike_report`	Generate stakeholder report

Integration Patterns¶

VS Code + Copilot¶

Install Klondike MCP server:
```
klondike mcp install
```
Copilot Chat will have access to Klondike tools
Use natural language:
"What features are in progress?"
"Start working on F001"
"End my session with a summary"

Custom AI Agent¶

# Example: Using Klondike MCP with a custom agent
import subprocess
import json

def get_status():
    result = subprocess.run(
        ["klondike", "status", "--json"],
        capture_output=True,
        text=True
    )
    return json.loads(result.stdout)

def start_feature(feature_id):
    subprocess.run(["klondike", "feature", "start", feature_id])

# In your agent loop
status = get_status()
print(f"Project: {status['project_name']}")
print(f"Completion: {status['completion']}%")

CI/CD Integration¶

# GitHub Actions example
- name: Check Klondike Status
  run: |
    klondike validate
    klondike status

- name: Verify Features
  run: |
    # Verify all in-progress features have evidence
    klondike feature list --status verified --json > verified.json

Configuration¶

MCP Config File¶

The MCP config file (mcp-config.json) structure:

{
  "servers": {
    "klondike": {
      "command": "klondike",
      "args": ["mcp", "serve"],
      "transport": "stdio"
    }
  }
}

VS Code Settings¶

The MCP server is configured in VS Code settings:

{
  "mcp.servers": {
    "klondike": {
      "command": "klondike",
      "args": ["mcp", "serve"]
    }
  }
}