Claude MCP Stock Screener: StockMarketScan in Claude, Cursor and Continue

What is MCP and why should you care

The Model Context Protocol is an open standard — originally proposed by Anthropic — that lets any LLM client talk to any external tool or data source through a consistent interface. Think of it as the USB of AI tooling. Claude Desktop, Cursor, Continue, Cline, Windsurf, Zed, and dozens of other clients now ship MCP support out of the box. When a client is connected to an MCP server, the model sees a list of tools it can call, and it can decide, based on what you ask, which tool to invoke and how.

For a platform like StockMarketScan, this is the cleanest way to put our data in front of your model without locking you into any single chat product. You pick the client you already use. We just register as a tool source.

What a Claude MCP stock screener gives you — 20 tools in one server

The StockMarketScan MCP server exposes every core capability of the platform as a named tool:

• Screeners (3 tools). List all 24 curated screeners, pull the live rows of any specific one, and run confluence queries across multiple screeners at once (which stocks are in both Golden Cross and Hot Prospects today?).
• Chart Patterns (2 tools). Ask for every detected pattern on a single symbol, or search across a set of screeners for all stocks currently forming a specific pattern (find me hot prospects with a cup and handle).
• Options Flow (5 tools). Daily options flow overview with call/put ratios, consecutive-day streaks, and premium; historical timeline for any single symbol; curated high-conviction signals filtered by screener confluence; and contract-level unusual options activity.
• Stock Info (3 tools). Fundamentals and quote data for any US symbol, plus raw OHLCV candles.
• Composite (2 tools). get_stock_report fans out to multiple endpoints in parallel and returns a full research brief. search_setups ranks the strongest bullish or bearish opportunities across the entire signal universe.
• Market Context (3 tools). Daily market momentum (advancers/decliners, new highs/lows), Trend Monitor output, and ticker-to-ticker connections.
• Education (1 tool). Plain-language explanations of platform-specific terms like strength_score, vol_oi_ratio, consecutive_days.

And one ping tool for liveness checks, bringing the total to 20.

What you can actually do with it

A few real examples that work today, straight out of the box, in any MCP-capable client:

"Which stocks appear in both hot-prospects and golden-cross today, and do any of them have a bullish chart pattern forming?"

The model calls search_stocks_in_screeners to find the confluence, then search_patterns with the resulting symbols to filter for bullish patterns, then summarizes the list for you.

"Show me AMD's options flow history for the last 30 days and tell me when the sentiment flipped."

The model calls get_options_flow_timeline for AMD with a 30-day window, reads back the daily rows, and describes when the call/put ratio crossed above or below 1.0.

"Build a full stock report for NVDA — screeners it appears in, chart patterns, options flow state, and current market context."

The model calls get_stock_report, which internally fans out to six of our endpoints in parallel and returns a structured bundle. The model then narrates it in plain English.

"What are today's strongest bullish options flow signals, and which ones have large-cap liquidity?"

The model calls get_options_flow_signals to get the curated list, then cross-references each symbol's fundamentals using get_stock_info, and filters the result by market cap.

These are not demo tricks. They are queries that would otherwise take you ten browser tabs and a spreadsheet. The model handles the orchestration, and we handle the data.

How authentication works — Bring Your Own Key

The StockMarketScan MCP server is designed around a very simple principle: you own your API key, we don't.

When you connect Claude Desktop or Cursor to the server, the client passes your personal sms_* API key in every request as the X-API-Key header. The server validates it once per session and then proxies each tool call to api/v1/* on your behalf. Your rate limits, quota, and tier are entirely your own. You can rotate the key from the Settings page at any time and the server picks it up on the next connection.

This is called BYOK — Bring Your Own Key — and it means we do not hold a long-lived secret tied to your account, we cannot run queries you did not initiate, and nothing is ever shared across users. Each MCP session is its own little sandbox.

Installation — three clients, three configs

Claude Desktop. Edit claude_desktop_config.json (on macOS at ~/Library/Application Support/Claude/claude_desktop_config.json) and add the following under mcpServers:

{
  "mcpServers": {
    "stockmarketscan": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.stockmarketscan.com/mcp"],
      "headers": {
        "X-API-Key": "sms_your_key_here"
      }
    }
  }
}

Restart Claude Desktop. You will see stockmarketscan in the tool picker, and the model can now call any of the 20 tools mid-conversation.

Cursor. Go to Settings → Features → Model Context Protocol → Add New MCP Server. Paste:

{
  "name": "stockmarketscan",
  "url": "https://mcp.stockmarketscan.com/mcp",
  "headers": { "X-API-Key": "sms_your_key_here" }
}

Continue (VS Code / JetBrains). Edit ~/.continue/config.json and add an entry under experimental.modelContextProtocolServers with transport type sse, URL https://mcp.stockmarketscan.com/mcp, and the same header.

The full config snippets, including examples for other clients, are in the install guide.

Which plan do I need

The MCP server is available on Basic and Pro plans. Basic gives you 15 requests per minute and 500 per day across all screener, pattern, and market tools. Pro raises that to 30 per minute and 2,000 per day and unlocks the four options flow tools. The rate limits apply per API key, not per MCP session, so you can keep the client connected all day and still get the same budget.

If you do not yet have an account, you can start on the Free plan — the MCP server itself will not work without a Basic or Pro key, but the rest of the site is fully functional so you can explore before subscribing.

Where we go from here

This is the first version. It already covers every core capability of the platform, but we are not done. A few things on our list for the next iterations:

• Streaming prompts. Pre-built prompt packs that get installed alongside the server, so you can type /options-flow-report NVDA in Claude Desktop and get a structured answer.
• More context tools. Earnings calendar, economic calendar, sector performance, insider activity.
• Better caching. The server already does per-session in-memory caching for the heavier endpoints; we want to move to a shared cache so the first query for today's screener gets cached for everyone.
• Multi-symbol tool calls. Right now get_stock_info is called per symbol. We will add batch variants to reduce the number of round-trips the model has to make.

If there is something specific you want from the server — a tool that doesn't exist yet, a rate limit that doesn't fit your workflow, a client that is awkward to configure — tell us. MCP is a two-year-old protocol and the ecosystem is still finding its shape. We would rather build exactly what power users need than guess.

Get started

1. Generate an API key at stockmarketscan.com/settings
2. Paste one of the configs above into your client
3. Ask your first question

If you get stuck, the full install guide with all client configs is at stockmarketscan.com/mcp. Welcome to a much shorter feedback loop between your questions and the market data that answers them.