TellWorldAI World Archive
Ask中文

Home›Practice›Practice

Practice

Deploying MCP servers

An MCP server can run on your own machine or be deployed as a container app in the cloud. This guide covers the local steps, the Azure Container Apps sample, and why SSE endpoints are now legacy.

Who this is for and what you will be able to do

An MCP server that only runs on your own computer cannot be used by anyone else or from another environment. Deployment makes its tools and resources reachable beyond your machine. By the end of this guide you will be able to tell local use from remote deployment, follow the sample to deploy a server to Azure Container Apps, and recognize why /sse endpoints belong to an older transport [1].

Check the transport first: SSE is legacy

Configuration examples that use a /sse endpoint target the legacy HTTP+SSE transport. MCP 2026-07-28 remote servers use Streamable HTTP, normally at an endpoint the server defines, such as /mcp [1]. If you are building a new remote server, choose the transport by that version of the spec, and treat /sse examples as older material [1].

Running locally

If the server runs on the user's own machine, download the server first if you did not write it, then start the server process [1]. Only SSE servers need the networking steps. stdio servers do not. For an SSE server, make sure it is reachable on the expected port and connect clients with a local URL such as http://localhost:3000 [1].

Cloud options

MCP servers can be deployed to several kinds of cloud platform [1]:

  • Serverless functions, for lightweight servers.
  • Container services, such as Azure Container Apps, AWS ECS or Google Cloud Run.
  • Kubernetes, for running and managing servers in a cluster for high availability.

Example: deploying to Azure Container Apps

Azure Container Apps supports MCP servers, but this support is still a work in progress, and it currently supports SSE servers only [1]. The sample lives in its own repository [2].

1. Clone the sample

Clone the repository [1][2]:

git clone https://github.com/anthonychu/azure-container-apps-mcp-sample.git

2. Run it locally first

Create the environment and install dependencies with uv venv and uv sync. Then set the API_KEYS environment variable, using export on linux or macOS and set on Windows. Finally start the server in FastAPI's development mode [1]:

uv venv
uv sync

# linux/macOS
export API_KEYS=<AN_API_KEY>
# windows
set API_KEYS=<AN_API_KEY>

uv run fastapi dev main.py

3. Connect from VS Code

Create an mcp.json file in a .vscode directory with the content below [1]:

{
    "inputs": [
        {
            "type": "promptString",
            "id": "weather-api-key",
            "description": "Weather API Key",
            "password": true
        }
    ],
    "servers": {
        "weather-sse": {
            "type": "sse",
            "url": "http://localhost:8000/sse",
            "headers": {
                "x-api-key": "${input:weather-api-key}"
            }
        }
    }
}

The key is requested through a password input named weather-api-key, and the file refers to it as ${input:weather-api-key} rather than storing it directly. Once the SSE server is running, click the play icon in the JSON file. GitHub Copilot should then list the server's tools, shown by a Tool icon [1].

4. Deploy

Run this command to deploy [1]:

az containerapp up -g <RESOURCE_GROUP_NAME> -n weather-mcp --environment mcp -l westus --env-vars API_KEYS=<AN_API_KEY> --source .

The command takes a resource group (-g), an app name (-n), an environment (--environment), a location (-l) and environment variables (--env-vars). --source . deploys from the current folder [1].

Limits and common questions

  • The sample supports only SSE servers, so its endpoints and configuration follow the older transport [1].
  • The deploy command uses westus as the location and placeholder values for the resource group and API key. Replace them with your own [1].
  • An article from Microsoft's Tech Community covers hosting remote MCP servers in Azure Container Apps [3].

Next step

The next topic in this series is advanced server topics [1]. The sample repository and the Azure article are listed above [2][3].

Adapted from the deployment lesson of Microsoft's open-source course *mcp-for-beginners*[1], released under the MIT License.

Sources

  1. 1
    Deploying MCP ServersMicrosoft mcp-for-beginners (MIT)
  2. 2
    Azure Container Apps MCP sampleazure-container-apps-mcp-sample (GitHub)
  3. 3
    Azure Container Apps articleMicrosoft Tech Community

Cite this page: Deploying MCP servers. TellWorld, https://www.tellworld.net/practice/practice-mcp-deployment. Updated 2026-11-14

Related news

View all