Point Claude at the stock market

HoodGrow runs a hosted MCP server over its Robinhood Chain stock-token data. Connecting Claude Code to it is one command:

claude mcp add --transport http hoodgrow https://www.hoodgrow.com/api/mcp

That's the whole setup. Every tool is read-only, so anonymous access just works — rate-limited per IP, with a free API key available when you want more headroom. Claude Desktop and any other MCP client that speaks Streamable HTTP connect with the same URL.

Once connected, the assistant sees ten tools and picks them itself. Things you can just ask:

"What moved on Robinhood Chain today?" — top gainers, losers, volume and liquidity leaders.
"What's NVDA's real supply after the split?" — corporate-action-adjusted supply (ERC-8056), not a naive totalSupply() read.
"Show me recent whale trades." — the large-trade feed indexed from the DEX pools, with sizes and transaction hashes.
"Get 90 days of GME daily candles." — OHLC history with per-candle volume, ready for backtesting.
"What would a $10k AAPL buy cost in slippage?" — per-pool price-impact estimates.

Prices come from Chainlink feeds read on-chain and refresh every 15 minutes; every response carries its own timestamp, so the assistant knows exactly how fresh its numbers are.

The hosted server is the zero-friction path. For heavier or fully autonomous use, the same tools ship as an npm package — npx hoodgrow-mcp — where the caller brings its own access: a free API key, a prepaid credit balance, or an x402 wallet that pays per call in USDC with no account anywhere — here's how agents pay for that. TypeScript and Python SDKs cover the no-MCP case.

The MCP server reads the same snapshot database every HoodGrow page renders from — an assistant and a visitor looking at the same token can never disagree. Why that data layer exists at all is its own post: Why HoodGrow?