How to build an AI Agent with Claude Agent SDK and the ClickHouse MCP Server

In this guide you'll learn how to build a Claude Agent SDK AI agent that can interact with ClickHouse's SQL playground using ClickHouse's MCP Server.

Example notebook

This example can be found as a notebook in the examples repository.

Prerequisites

• You'll need to have Python installed on your system.
• You'll need to have pip installed on your system.
• You'll need an Anthropic API key

You can run the following steps either from your Python REPL or via script.

Install libraries

Install the Claude Agent SDK library by running the following commands:

!pip install -q --upgrade pip
!pip install -q claude-agent-sdk
!pip install -q ipywidgets

Setup credentials

Next, you'll need to provide your Anthropic API key:

import os, getpass
os.environ["ANTHROPIC_API_KEY"] = getpass.getpass("Enter Anthropic API Key:")

Enter Anthropic API Key: ········

Next, define the credentials needed to connect to the ClickHouse SQL playground:

env = {
    "CLICKHOUSE_HOST": "sql-clickhouse.clickhouse.com",
    "CLICKHOUSE_PORT": "8443",
    "CLICKHOUSE_USER": "demo",
    "CLICKHOUSE_PASSWORD": "",
    "CLICKHOUSE_SECURE": "true"
}

Initialize MCP Server and Claude Agent SDK agent

Now configure the ClickHouse MCP Server to point at the ClickHouse SQL playground and also initialize our agent and ask it a question:

from claude_agent_sdk import query, ClaudeAgentOptions, AssistantMessage, UserMessage, TextBlock, ToolUseBlock

options = ClaudeAgentOptions(
    allowed_tools=[
        "mcp__mcp-clickhouse__list_databases",
        "mcp__mcp-clickhouse__list_tables", 
        "mcp__mcp-clickhouse__run_select_query",
        "mcp__mcp-clickhouse__run_chdb_select_query"
    ],
    mcp_servers={
        "mcp-clickhouse": {
            "command": "uv",
            "args": [
                "run",
                "--with", "mcp-clickhouse",
                "--python", "3.10",
                "mcp-clickhouse"
            ],
            "env": env
        }
    }
)


async for message in query(prompt="Tell me something interesting about UK property sales", options=options):
    if isinstance(message, AssistantMessage):
        for block in message.content:
            if isinstance(block, TextBlock):
                print(f"🤖 {block.text}")
            if isinstance(block, ToolUseBlock):
                print(f"🛠️ {block.name} {block.input}")
    elif isinstance(message, UserMessage):
        for block in message.content:
            if isinstance(block, TextBlock):
                print(block.text)

Note the code inside the for block is filtering the output for brevity.

🤖 I'll query the ClickHouse database to find something interesting about UK property sales.

Let me first see what databases are available:
🛠️ mcp__mcp-clickhouse__list_databases {}
🤖 Great! There's a "uk" database. Let me see what tables are available:
🛠️ mcp__mcp-clickhouse__list_tables {'database': 'uk'}
🤖 Perfect! The `uk_price_paid` table has over 30 million property sales records. Let me find something interesting:
🛠️ mcp__mcp-clickhouse__run_select_query {'query': "\nSELECT \n    street,\n    town,\n    max(price) as max_price,\n    min(price) as min_price,\n    max(price) - min(price) as price_difference,\n    count() as sales_count\nFROM uk.uk_price_paid\nWHERE street != ''\nGROUP BY street, town\nHAVING sales_count > 100\nORDER BY price_difference DESC\nLIMIT 1\n"}
🤖 Here's something fascinating: **Baker Street in London** (yes, the famous Sherlock Holmes street!) has the largest price range of any street with over 100 sales - properties sold for as low as **£2,500** and as high as **£594.3 million**, a staggering difference of over £594 million!

This makes sense given Baker Street is one of London's most prestigious addresses, running through wealthy areas like Marylebone, and has had 541 recorded sales in this dataset.