The domain allowlist and the minimum-score floor both cut noise before it reaches the model, which matters for a query type where the low-quality results outnumber the good ones. Both symbol and company name go into the query, so a ticker that is also a common word does not derail it. One key, one provider, and Tavily bills you for usage.
An MCP server that turns a stock symbol and company name into a news search against the Tavily API. The query templates are built for the finance case, so you get results about the company rather than everything matching the ticker string.
- News for a ticker, addressed by `symbol` and `companyName`, with `maxResults` controlling how much comes back
- Search depth you can raise per query — `searchDepth` set to `advanced` for a deeper pass
- A relevance floor — `minScore` filters out weak matches before they reach the conversation
- Domain-specific filtering — `includeDomains` restricts results to sources you trust, such as reuters.com and bloomberg.com
- Multiple customizable search query templates, with type-safe operations in TypeScript
A Tavily API key, set as `TAVILY_API_KEY`. Node.js v16 or higher. Run it with `npx -y search-stock-news-mcp@latest`, install through Smithery with `npx -y @smithery/cli install search-stock-news-mcp --client claude`, or add it from the Cline marketplace. In Cursor the command form is `env TAVILY_API_KEY=your-api-key-here npx -y search-stock-news-mcp@latest`. Package `search-stock-news-mcp` 1.0.12, MIT licensed.
One command plus a key — npx -y search-stock-news-mcp@latest, then supply credentials