MCP Scaleway Functions

Model Context Protocol (MCP) server to manage and deploy Scaleway Serverless Functions using the Model Context Protocol standard.

[!CAUTION]
This project is unofficial and not affiliated with or endorsed by Scaleway.
Some small safety measures are in place to prevent the LLM from doing destructive actions,
but they're not foolproof.
Use at your own risk.

Getting Started

Download the latest release from the releases page or build it from source using Go.

Run the MCP server:

./mcp-scaleway-functions

By default, the MCP server runs with the SSE transport on http://localhost:8080, but you can also change it to use Standard I/O (stdio) transport via the --transport stdio flag.

Then, configure your IDE or tool of choice to connect to the MCP server. Here are some examples:

VSCode (sse example)

Add a new server configuration in your .vscode/mcp.json file:

{
	"servers": {
		"mcp-scaleway-functions": {
			"url": "http://localhost:8080",
			"type": "http",
		}
	},
}

Crush (stdio example)

Crush is an open-source coding agent that supports MCP. You can find more information about in the Crush repository.

Add a new server configuration in your ~/.config/crush/crush.json file:

{
  "$schema": "https://charm.land/crush.json",
  "mcp": {
    "scaleway-functions": {
      "type": "stdio",
      "command": "mcp-scaleway-functions",
      "args": ["--transport", "stdio"],
      "timeout": 600,
      "disabled": false
    }
  }
}

You can even use Crush with Scaleway Generative APIs by adding a new provider in the same ~/.config/crush/crush.json file:

{
  "mcp": {
	// ... see above ...
  },
  "providers": {
    "scaleway": {
      "name": "Scaleway",
      "base_url": "https://api.scaleway.ai/v1/",
      "type": "openai",
	  // To fetch from environment variables, use the `$VAR_NAME` syntax.
	  // Note: this key requires the "GenerativeApisModelAccess" permission.
      "api_key": "$SCW_SECRET_KEY",
      "models": [
        {
          "name": "Qwen coder",
          "id": "qwen3-coder-30b-a3b-instruct",
          "context_window": 128000,
          "default_max_tokens": 8000
        }
      ]
    }
  }
}

That's it 🎉! Have fun vibecoding and vibedevoopsing as you please.

Configuration

By default, the MCP server reads from the standard Scaleway configuration file located at ~/.config/scw/config.yaml.

Further configuration can be done via the
Scaleway environment variables to configure the MCP server.

For instance, you can set a region to work in via the SCW_DEFAULT_REGION environment variable.

SCW_DEFAULT_REGION=nl-ams ./mcp-scaleway-functions

Available Tools

Tool	Description
`create_and_deploy_function_namespace`	Create and deploy a new function namespace.
`list_function_namespaces`	List all function namespaces.
`delete_function_namespace`	Delete a function namespace.
`list_functions`	List all functions in a namespace.
`list_function_runtimes`	List all available function runtimes.
`create_and_deploy_function`	Create and deploy a new function.
`update_function`	Update the code or the configuration of an existing function.
`delete_function`	Delete a function.
`download_function`	Download the code of a function. This is useful to work on an existing function.
`fetch_function_logs`	Fetch the logs of a function.
`add_dependency`	Add a dependency to a local function. Useful for dependencies that rely on native code and therefore need Docker to be installed.

Debugging

You can enable debug logging by using the --debug flag when starting the MCP server. This will log all requests and responses to/from the Scaleway API.

To configure the log level, use the --log-level flag (default is info). Available log levels are: debug, info, warn, error.

Logs are stored in the $XDG_STATE_HOME/mcp-scaleway-functions directory (usually ~/.local/state/mcp-scaleway-functions).

Development

Running tests:

go tool gotestsum --format testdox

Generating mocks:

go tool mockery