Back to Blog/devtools

How to Set Up Docker's MCP Gateway for Multiple Servers

Set up Docker MCP Toolkit profiles and connect a client to the Gateway, with an explicit stdio configuration and a practical verification procedure.

Adam BushAdam BushAugust 19, 20263 min read
#mcp#developer#docker#devtools

Docker's MCP Gateway provides a common connection for clients while orchestrating MCP servers. It does not mean that every server runs inside one shared container, as an earlier version of this article stated. Docker describes it as a proxy managing server configuration and credentials. Gateway overview.

Documentation reviewed September 10, 2026; no authenticated installation was performed for this update.

Start with Docker Desktop's Toolkit

Docker's current getting-started guide describes the profile interface in Desktop 4.62 and later. Enable MCP Toolkit, create a profile, and add one server from the catalog. Complete that server's required configuration before connecting a client. In the Clients tab, choose your application and use Connect. Earlier Desktop versions can present a different interface. Docker Toolkit setup.

Choose a server for which you already have a simple read-only task. A profile containing many tools makes the first failure harder to attribute.

Manual stdio connection

If your client needs manual configuration, first create a profile named mcpfind and configure a server in it. This VS Code example starts the gateway using that profile:

json
{
  "servers": {
    "MCP_DOCKER": {
      "type": "stdio",
      "command": "docker",
      "args": ["mcp", "gateway", "run", "--profile", "mcpfind"]
    }
  }
}

Save it in .vscode/mcp.json. The servers key and type field follow VS Code's configuration reference; other clients can require a different top-level key. Docker documents the --profile connection pattern in its Toolkit setup.

The Docker command must be available to the client. For an Engine-only installation, follow Docker's separate Gateway plugin installation instructions before trying docker mcp. Manual Gateway installation.

Verify the server behind the gateway

Inspect the client connection, then request a known item using the server you added. For example, with a properly authorized repository server, ask for the title and link of an issue you can open yourself. Check both the tool result and the source page.

If no tools appear, check the profile name and required server configuration. If tools appear but access fails, investigate that server's credentials and permissions. A working gateway process does not establish that every upstream service is authenticated.

Decide whether the extra layer helps

A gateway is useful when one profile makes several required connections easier to maintain. For a single hosted integration, direct client setup can be simpler. Compare the number of moving parts you need to keep running, not the number of servers you could theoretically enable.

Use MCP Find's directory for discovery, then confirm catalog support in Docker before expecting a listing to work through the Toolkit. A general directory listing is not a Docker compatibility certification.

Frequently Asked Questions

Does the Gateway put all servers in one container?

No. It orchestrates and proxies server connections. Do not assume that a gateway connection means all server processes share one container.

Can I use Docker Engine without Desktop?

Docker documents installing the MCP Gateway separately as a Docker CLI plugin for Engine-only environments.

Should I enable every server in a catalog?

Start with only the servers needed for your task. Add another after you have verified the first connection and its results.

Related Articles