MCP server: Deployment experience server connection prompt

Ask summary

When customers run the Azure Functions: Deploy to Function app or Azure Functions: Deploy to Azure commands, detect whether the app is an MCP server, and if yes, add the following prompt:

Connect to remote MCPserver post deployment

• Yes
• No

If selected Yes

When deployment completes, create (if needed) and open the mcp.json file with server registration information.

• Access key is required by default, so the configuration should be:

  "servers": {
      "remote-mcp-server": {
          "type": "http",
          "url": "<depends on MCP server flavor>",
          "headers": {
              "x-functions-key": "${input:command to get key}"
          }
      }
  },
  "inputs": [
      {
          "type": "promptString",
          "id": "functions-key",
          "description": "Functions App Key",
          "password": true
      }
  ]
  

• DON'T run the server directly, but populate the url (to provide users the chance to correct the url if needed)
  • Self-hosted MCP server url: retrieve the app's root url and append /mcp
  • MCP extension server url: retrieve the app's root url and append /runtime/webhooks/mcp
• x-functions-key:
  • MCP extension servers need the Systems MCP extension key
  • Self-hosted servers need the default key

If selected No

Do nothing.

Other requirements

• When users start server, retrieve and inject the function key automatically by running a command
• Show this prompt only if the server is not already registered in the mcp.json
• Don't show prompt again if customer has already answered prompt

E2E flow summary

1. Customers run the Azure Functions: Deploy to Function app or Azure Functions: Deploy to Azure
2. Select function app or Create new function app (existing prompt today)

If selects function app
3. --> This ask: Connect to remote server in VSCode workspace
4. Pop up to confirm deployment (exists today)

If selects Create new function app
3. Enter app name (existing prompt)
4. Select location for new resources (existing prompt)
5. Select run time stack (existing prompt)
6. Select resource authentication type (existing prompt)
7. --> This ask: Connect to remote server in VSCode workspace

No files, tests, or entry points are named. Start by tracing the Azure Functions deployment commands and their existing prompts, then identify how the workspace mcp.json is created or opened; done means the prompt, URL and key configuration, registration checks, and remembered responses work for both deployment flows.