How to Query Financial Data with Anthropic's MCP and Claude Sonnet-4

In this tutorial, we'll explore how to leverage Anthropic's Model-Context Prompt (MCP) feature with Claude Sonnet-4, integrated with CoinAPI, to query financial data from multiple sources.. MCP (Model Context Protocol) allows you to connect Claude to external APIs, enabling powerful, context-aware queries. We'll walk through setting up a Python environment, configuring credentials, and querying two financial data sources to retrieve and summarize Apple's SEC 8-K filings.

---

Contents

1. Install Prerequisites

2. Configure Credentials

3. One-Shot Query with Two MCP Servers

4. Continuing the Conversation

5. Mixing Structured and Natural-Language Queries

1. Install Prerequisites

To get started, install the required Python packages: anthropic for interacting with Claude and rich for pretty-printing responses.

pip install -q anthropic rich

2. Configure Credentials

You'll need an Anthropic API key from console.anthropic.com. For security, store it in an environment variable to keep it out of your code.

import os
os.environ["ANTHROPIC_API_KEY"] = "YOUR_ANTHROPIC_API_KEY"

Replace YOUR_ANTHROPIC_API_KEY with your actual key.

3. One-Shot Query with Two MCP Servers

We'll query two external data sources using MCP:

• Historical stock-price feed (mcp_api_apibricks_io)

• SEC filings stream (mcp_sec_apibricks_io)

The MCP configuration is passed as a list of dictionaries to the mcp_servers parameter. Below, we ask Claude to find the date of Apple's most recent 8-K filing.

from anthropic import Anthropic
from rich import print

client = Anthropic()

api_key = "YOUR-FINFEED-API-KEY"

MCP_SERVERS = [
    {
        "type": "url",
        "url": "https://mcp-historical.stock.finfeedapi.com/mcp",
        "name": "mcp_api_apibricks_io",
        "authorization_token": api_key
    },
    {
        "type": "url",
        "url": "https://mcp.sec.finfeedapi.com/sse",
        "name": "mcp_sec_apibricks_io",
        "authorization_token": api_key
    }
]

response = client.beta.messages.create(
    model="claude-sonnet-4-20250514",
    max_tokens=1000,
    messages=[{
        "role": "user",
        "content": "When did Apple last file an 8-K report?"
    }],
    mcp_servers=MCP_SERVERS,
    extra_headers={"anthropic-beta": "mcp-client-2025-04-04"}
)

print(response.content)

Output Explanation: Claude queries the SEC filings stream and returns a structured response, indicating that Apple's most recent 8-K was filed on May 12, 2025, covering items 8.01 and 9.01. It also provides a brief history of recent filings, showcasing how MCP integrates data from external sources seamlessly.

---

4. Continuing the Conversation

To maintain context, append the previous user prompt and assistant response to the messages list. Here, we ask Claude to summarize the key points of the May 12, 2025, 8-K filing in bullet form.

conversation = [
    {"role": "user", "content": "When did Apple last file an 8-K report?"},
    {"role": "assistant", "content": response.content}
]

conversation.append({
    "role": "user",
    "content": "Great, could you summarize the key points from that 8-K in bullet form?"
})

follow_up = client.beta.messages.create(
    model="claude-sonnet-4-20250514",
    max_tokens=750,
    messages=conversation,
    mcp_servers=MCP_SERVERS,
    extra_headers={"anthropic-beta": "mcp-client-2025-04-04"}
)

print(follow_up.content)

Output Summary:

• Total Amount Raised: $4.5 billion through four series of corporate bonds

• Bond Details:

  • $1.5 billion in 4.000% Notes due 2028

  • $1.0 billion in 4.200% Notes due 2030

  • $1.0 billion in 4.500% Notes due 2032

  • $1.0 billion in 4.750% Notes due 2035

• Interest Payment Schedule: Semi-annual payments on May 12 and November 12, starting November 12, 2025

• Underwriters: Goldman Sachs, Barclays Capital, BofA Securities, and J.P. Morgan Securities

• Security Type: Senior unsecured obligations

• Legal Structure: Issued under an existing indenture with Bank of New York Mellon Trust Company

• Purpose: Likely a routine debt financing transaction

This demonstrates Claude's ability to maintain conversational context and extract detailed insights from the same data source.

---

5. Mixing Structured and Natural-Language Queries

Claude can handle structured outputs, such as JSON, for precise data extraction. Let's request only the filing date and a JSON summary of the 8-K filing.

conversation.append({
    "role": "assistant",
    "content": follow_up.content
})

conversation.append({
    "role": "user",
    "content": "Return ONLY a JSON object with keys `filing_date` and `highlights`."
})

json_reply = client.beta.messages.create(
    model="claude-sonnet-4-20250514",
    max_tokens=300,
    messages=conversation,
    mcp_servers=MCP_SERVERS,
    extra_headers={"anthropic-beta": "mcp-client-2025-04-04"}
)

print(json_reply.content)

Output:

{
  "filing_date": "2025-05-12",
  "highlights": [
    "Issued $4.5 billion in corporate bonds across four series",
    "$1.5 billion in 4.000% Notes due 2028",
    "$1.0 billion in 4.200% Notes due 2030",
    "$1.0 billion in 4.500% Notes due 2032",
    "$1.0 billion in 4.750% Notes due 2035",
    "Underwritten by Goldman Sachs, Barclays Capital, BofA Securities, and J.P. Morgan Securities",
    "Semi-annual interest payments beginning November 12, 2025",
    "Senior unsecured obligations ranking equally with other Apple debt"
  ]
}

This JSON output shows Claude's flexibility in delivering structured data, ideal for developers integrating with other systems.

---

Conclusion

Using Anthropic's MCP with Claude Sonnet-4, you can seamlessly query multiple data sources, maintain conversational context, and generate both natural-language and structured outputs. This tutorial demonstrated how to:

• Set up the Anthropic SDK and configure credentials securely

• Query financial data from two MCP servers

• Continue a conversation with context

• Extract structured JSON responses

Try experimenting with other MCP-compatible APIs or different queries to unlock the full potential of Claude's multi-context capabilities!

Tags: #Anthropic #Claude #MCP #Python #FinancialData #API