How to use sample in mcp server to utilize in adk?

[shivsant]

fastmcp supports sample, but when I use adk, it says ValueError: Client does not support sampling

This is an example of how to use sampleing in fastmcp.

from fastmcp import FastMCP, Context

mcp = FastMCP("SamplingDemo")

@mcp.tool
async def analyze_sentiment(text: str, ctx: Context) -> dict:
"""Analyze the sentiment of text using the client's LLM."""
prompt = f"""Analyze the sentiment of the following text as positive, negative, or neutral.
Just output a single word - 'positive', 'negative', or 'neutral'.

Text to analyze: {text}"""

# Request LLM analysis
response = await ctx.sample(prompt)

# Process the LLM's response
sentiment = response.text.strip().lower()

# Map to standard sentiment values
if "positive" in sentiment:
    sentiment = "positive"
elif "negative" in sentiment:
    sentiment = "negative"
else:
    sentiment = "neutral"

return {"text": text, "sentiment": sentiment}

[lizzij]

[Workaround]: Use Server-Controlled Mode (using the Server's LLM)

# Set up server's private LLM handler
llm_handler = OpenAISamplingHandler(
    default_model="gpt-4o",  # Specify the gpt-4o model
    client=OpenAI(
        api_key=API_KEY,
    ),
)


# Create the FastMCP Server instance
mcp = FastMCP(
    name="SamplingDemo",

    # This tells the server to use its own LLM
    sampling_handler=llm_handler,

    # This forces the server to *always* use its handler
    # and never ask the client (your ADK agent)
    sampling_handler_behavior="always"
)

[surajksharma07]

@shivsant lizzij's suggestion above (giving the FastMCP server its own sampling handler) is one way around it but worth knowing ADK also added proper support for this since. McpToolset now takes a sampling_callback so your agent's own LLM can answer the server's ctx.sample() calls directly, no server-side handler needed. Closing this out since it looks addressed, feel free to reopen or comment if you're still running into it.