Real-world market data for your agents
444 market data tools behind one MCP endpoint. Here is what agents ask them.
A model knows what the web said when it was trained. Prowl tells it what the market says today. Connect Claude Code, Cursor, Codex or your own stack to https://prowl.chat/mcp and the agent can call search, ads, SEO, app store, review, social, trends and finance data on its own – or ask Prowl Agent on prowl.chat, which runs on the same tools. Each page below names the tools a use case calls and what they do not cover.
By use case
- Market research — Competitors, demand and market size from SERPs, trends, reviews and ads.
- SEO & AI search — Rankings, keywords, backlinks from four indexes, and mentions inside AI answers.
- Marketing & ads — Live ads from the Meta, TikTok, LinkedIn and Google ad libraries, plus keyword CPCs.
- Social media — YouTube, Instagram, TikTok, Facebook, Reddit and Pinterest profiles, posts and engagement.
- Creative — Long-running ads, hooks and formats that already work, across platforms.
- Trends — Google Trends, clickstream trends, news and regional interest over time.
- Investing & trading — Google Finance quotes, market movers and tickers, with the news and search interest around them.
- Product development — App Store and Google Play rankings and reviews, feature complaints, patents and papers.
- E-commerce — Amazon, Walmart and Google Shopping products, prices, sellers and reviews.
- Local & promotion — Google Maps and business profiles, local rankings, listings and reviews.
- Analytics — Traffic estimates, keyword volume and clickstream for any domain.
Connect your client
- Claude MCP server — Connect Prowl as an MCP server in Claude Code or Claude Desktop and Claude gains 444 live market-data tools, billed per call from a USD wallet.
- Cursor MCP server — Add Prowl to .cursor/mcp.json and Cursor gains 444 live market-data tools over MCP, billed per call from a USD wallet. No seats.
How every use case runs
Your agent finds a tool with prowl_search_tools, reads its schema with prowl_tool_info and calls it with prowl_call_tool. One call returns one answer and is billed on its own; the response carries the exact amount charged. When one call is not enough, prowl_analyze plans many calls, checks its numeric claims against independent tools and returns a report that says what it could not confirm.
The endpoint is https://prowl.chat/mcp over Streamable HTTP, authenticated with an Authorization: Bearer prowl_<your_key> header. No SDK, no per-provider keys, nothing to redeploy when the catalogue grows.
What you need to start
An MCP client and an API key – or no client at all, if you ask Prowl Agent on the site. New accounts get $15 in credit, enough for a first basic report or many single calls. The Getting Started walkthrough covers the wiring.