mirako-ai/mcp-remote-demo-server

Mirako mcp remote tools server demo

This project contains 3 demos:

1. locker

• This is a simple demo that offers a tool called get_available_lockers_from_zone , with a parameter zone required.
• The result of this demo is hardcoded for simplicity

2. weather

• This is a demo which fetch the weather data from api.weather.gov
• There are 2 tools provided get_alerts get_forecast
• This API only works for querying weather of US locations

3. show_photo

• This is a demo that offers a tool called show_photo demonstrating an image response
• The result of this demo is hardcoded for simplicity

Authentication

The auth_token specified in profile will be included in the header of ALL API calls as:

Authorization: Bearer [auth_token]

you may use the AuthTokenAuthProvider class from as your AuthProvider like this:

from auth.auth_provider import AuthTokenAuthProvider

mcp = FastMCP(name=__name__, auth=AuthTokenAuthProvider([auth_token]))

Profile ID

profile ID will be included in the header, you may call the get_profile_id(ctx) method in the samples to get the profile ID if needed

Supported transport options

streamable-http

Supported MCP Components

Tool

@mcp.tool()
def addition(
    a: Annotated[int, Field(description="first number to be added")],
    b: Annotated[int, Field(description="second number to be added")],
    ctx: Context
) -> dict:
    """Add 2 numbers"""
    # profile_id = get_profile_id(ctx)
    return a+b

Return type

Supported return types:

• str
  • return the response directly as str for text responses

Operation

Return a dict for operations, the operation name should be specified in operation key

Show Image

You should return an publicly accessible URL in image_url in order to show an image, format as below:

return {"operation": "show_image", "image_url": "https://your-domain.com/image.jpg"}

How to setup your tool in profile

You may setup multiple MCP tool servers in a profile

[
  {
    "url":"http://localhost:8080/mcp/",
    "auth_token":"abc"
  },
  {
    "url":"http://localhost:8081/mcp/",
    "auth_token":"abcd"
  }
]

Testing

You may use tester/main.py to test your MCP server. Change the URL, Auth Token and profile ID here if needed.

transport = StreamableHttpTransport(
    "http://localhost:8080/mcp/",
    auth=BearerAuth("your auth token"),
    headers={"M-PROFILE-ID": "your profile ID"},
)

Run tester

python tester.main.py

CloudRun Sample

Refer to