Step 7: Explore MCP Tools in GitHub Copilot (Local Endpoint)¶

Now, connect GitHub Copilot Chat to your locally running Function App's MCP endpoint to test the tools you implemented.

Verify MCP Server Configuration:
- In VS Code, open the .vscode/mcp.json file
- You should see the local-snippy server already configured
- If you see a "Running" status with tools count (e.g., "✓ Running | Stop | Restart | 5 tools"), the server is already connected
- If not connected, select the Start button in the file (appears at the top of the JSON configuration)
View MCP Server Logs (optional but helpful):
- Open the Command Palette by selecting Ctrl+Shift+P (Windows/Linux) or Cmd+Shift+P (macOS) or View > Command Palette from the toolbar
- Enter and select MCP: List Servers
- Select the local-snippy server from the dropdown
- Select Show Output to see the server logs
- Tip: Select the settings icon next to "MCP: local-snippy" in the output panel to change log level to "trace" for more detailed logs
Open GitHub Copilot Chat:
- Select Ctrl+Alt+I (Windows/Linux) or Cmd+Alt+I (macOS) or select the Copilot Chat icon in the Activity Bar
Configure Copilot Chat for Tools:
- Ensure Agent mode is active (select from the dropdown next to the model selector, or Ctrl+. / Cmd+.)
- At the bottom of the chat panel, select the Tools icon
- Make sure MCP Server: local-snippy and all its tools are checked
- Select Escape or select OK to confirm your selection
Troubleshooting MCP Server Connection Issues

If you're unable to connect to MCP servers, you may be encountering one of these issues:

Issue 1: Cache-related problems

To resolve cache issues:
1. Open VS Code command palette (Ctrl+Shift+P or Cmd+Shift+P).
2. Enter MCP Tools: Reset Cached Tools to execute the command.
Issue 2: Organization policy restrictions

If your GitHub Enterprise organization has disabled MCP servers via policies, you have two options:

Option A: Use a different GitHub account that isn't subject to the same restrictions

Option B: Use the MCP Inspector tool
1. Run the inspector locally: npx @modelcontextprotocol/inspector.
2. Open your browser and navigate to http://localhost:6274.
3. Test your MCP tools from this local interface.
These workarounds will let you complete the tutorial exercises even if organizational policies restrict direct MCP server access.
Test the save_snippet Tool:
- Open any code file (e.g., src/durable_agents.py)
- Select some interesting code sections (or it'll take the entire file as a snippet, which is fine, as well)
- In Copilot Chat, enter:
```
#local-snippy save this snippet as ai-agents-service-usage
```
  Press Enter or Send - If prompted by Copilot to use the save_snippet tool, select Allow
Test the get_snippet Tool:
- In Copilot Chat, enter:
```
#local-snippy get the snippet named ai-agents-service-usage
```
  Press Enter or Send - Copilot will suggest using the get_snippet tool - Select Allow
Experiment with Advanced AI Agent Tools:
- Try these prompts (no need to select code first):
  - generate a deep wiki for all snippets and place in a new file deep-wiki.md in project root (uses the deep_wiki tool)
  - create a code style guide based on the saved snippets and place in a new file code-style.md in project root (uses the code_style tool)
Note: These agent-based tools may take a minute to run as they orchestrate with configured tools and have self-reflection

Once they are done, open the files and Preview the generated Markdown (Ctrl+K V or Cmd+K V)
Check Function Logs:
- In the terminal where func start is running, you'll see logs for each tool invocation
- This confirms your MCP tools are working end-to-end

The ability to interact with your tools through natural language in Copilot demonstrates the power of the MCP protocol - AI assistants can seamlessly discover and use your Azure Functions without any custom integration code.