Claude MCP setup: connect Claude to real-time market data

To connect Claude to live market data, add TickerLayer as an MCP server. In Claude Code that is one claude mcp add command; in Claude Desktop it is a short config block; in your own app it is two parameters on the Claude API. Claude then sees ten read-only tools for quotes, bars, market hours and bond yields, and calls them instead of answering from memory.

MCP, the Model Context Protocol, is the open standard Claude uses to call external tools. The financial MCP server guide explains the protocol and the full tool catalog. This page is the practical half: setup in each Claude client, five prompts that test the connection, and what a good answer looks like in pre-market and after the close.

Pick a Claude client and get a key

• Claude CodeTerminal agent. Connects straight to the hosted server with your key as a header.
• Claude DesktopChat app. Uses a pinned local bridge from its config file; needs Node.js 18 or newer.
• Claude APIYour own app. The Messages API connects to the server for you with the MCP connector.

All three need a TickerLayer API key from the dashboard. A free API key works: the MCP server forwards each tool call to the REST API, so calls count against the free tier's 3,000 requests a month and follow your plan's entitlements.

Set up the Claude MCP server in Claude Code

1. 1Export your keyPut it in TICKERLAYER_API_KEY so it never appears literally in your shell history.
2. 2Add the serverRegister the hosted URL over HTTP with an x-api-key header.
3. 3Confirm it connectedList your MCP servers and check that tickerlayer reports connected.
4. 4Ask a market questionStart a session and ask something that needs live data, such as the status of the US market.

export TICKERLAYER_API_KEY="YOUR_API_KEY"

claude mcp add --transport http tickerlayer https://mcp.tickerlayer.com/mcp \
  --header "x-api-key: $TICKERLAYER_API_KEY"

claude mcp list

Set up Claude Desktop

Claude Desktop reads MCP servers from claude_desktop_config.json (Settings, Developer, Edit Config). The tested setup runs a pinned mcp-remote bridge that forwards to the hosted server over HTTP and keeps the key in an environment variable rather than in the command arguments. Paste this, replace the placeholder, and restart the app:

{
  "mcpServers": {
    "tickerlayer": {
      "command": "npx",
      "args": [
        "-y", "[email protected]", "https://mcp.tickerlayer.com/mcp",
        "--transport", "http-only",
        "--header", "x-api-key:${TICKERLAYER_API_KEY}"
      ],
      "env": { "TICKERLAYER_API_KEY": "YOUR_API_KEY" }
    }
  }
}

Connect from the Claude API

In your own application, the Messages API opens the MCP connection for you. Pass the server in mcp_servers and reference it from tools; the authorization_token is sent as a bearer token, which the TickerLayer server accepts as your API key.

import os

import anthropic  # pip install anthropic

client = anthropic.Anthropic()  # reads ANTHROPIC_API_KEY

response = client.beta.messages.create(
    model="claude-opus-5",
    max_tokens=16000,
    betas=["mcp-client-2025-11-20"],
    mcp_servers=[{
        "type": "url",
        "url": "https://mcp.tickerlayer.com/mcp",
        "name": "tickerlayer",
        "authorization_token": os.environ["TICKERLAYER_API_KEY"],
    }],
    tools=[{"type": "mcp_toolset", "mcp_server_name": "tickerlayer"}],
    messages=[{"role": "user", "content": "Is the US stock market open, and where is US:KO trading?"}],
)

for block in response.content:
    if block.type == "mcp_tool_use":
        print("tool:", block.name, block.input)
    elif block.type == "text":
        print(block.text)

Tool calls run on the API side and come back as mcp_tool_use and mcp_tool_result blocks alongside Claude's text. The AI trading bot tutorial extends this into a full agent loop with guardrails and a paper ledger.

Five prompts to test the connection

What a good answer looks like

On 28 September 2026 at 10:43 UTC, get_snapshot for US:KO returned a last price of 88.16 stamped 06:43 New York time, a previous close of 87.81 and a change of +0.35 (+0.40%), while get_market_status reported pre_market with is_open: true. The same numbers can produce an honest answer or a misleading one:

What happens when the market is closed

Ask about Tokyo in the European morning and the answer should change shape. At 10:52 UTC, 19:52 in Tokyo, the status tool reported the market closed and gave the next open:

{
  "asset": "stocks",
  "market": "JP",
  "status": "closed",1
  "is_open": false,
  "reason": "Outside primary session",
  "timezone": "Asia/Tokyo",
  "local_time": "2026-09-28T19:52:47.463+09:00",2
  "next_open": "2026-09-29T00:00:00.000Z"3
}

A snapshot still returns data at that hour, which is correct: it is the last known value. For JP:7203 it showed a last trade of 2,986.5 stamped 06:30:01 UTC, the 15:30 Tokyo close, against a previous close of 2,989.5. A good closed-market answer reads like this: "Tokyo is closed. JP:7203 last traded at 2,986.5 at the 15:30 close, down 3.0 (0.10%). Trading resumes at 09:00 Tokyo time, 00:00 UTC." The stock market hours guide covers sessions across markets.

ChatGPT: the same server over OAuth

ChatGPT apps and connectors are built on MCP as well, so the same server works there with an OAuth sign-in instead of a key. Developer mode is available on Plus, Pro, Business, Enterprise and Edu plans:

1. 1Enable developer modeIn ChatGPT on the web: Settings, Apps & Connectors, Advanced settings, Developer mode.
2. 2Create the connectorName it TickerLayer, set the MCP server URL to https://mcp.tickerlayer.com/mcp and choose OAuth.
3. 3Sign in onceChatGPT redirects to the TickerLayer sign-in; approve access.
4. 4Turn it on in a chatOpen the + menu, choose Developer mode and enable TickerLayer.

Troubleshooting the connection