Ask Aftershocks from your AI

Aftershocks has a free, read-only MCP server at https://aftershockshq.com/mcp: connect it to Claude, Cursor, ChatGPT or any other MCP client and ask for Coil scores, squeezes and Breaks in plain words. No account, no key.

Every answer names the hour it describes, the source (Coinbase public feed, closed hourly candles), what the score means and ends with "Not a forecast." Aftershocks measures volatility squeezes and breakouts. It never says which way price goes or what to buy or sell, and neither do these tools.

Claude

Claude Code, one line in a terminal:

claude mcp add --transport http aftershocks https://aftershockshq.com/mcp

Claude on the web or the desktop app: Settings, Connectors, Add custom connector. Name it Aftershocks and paste the address https://aftershockshq.com/mcp. No authentication.

Cursor

Add this to .cursor/mcp.json in your project, or to the global file in Cursor's settings:

{
  "mcpServers": {
    "aftershocks": { "url": "https://aftershockshq.com/mcp" }
  }
}

ChatGPT and other clients

Any client that speaks MCP over streamable HTTP takes the same address, https://aftershockshq.com/mcp. In ChatGPT it is a connector added under Settings with developer mode on.

The four tools

• get_coil: one coin by symbol or name (BTC, ETH-USD, Bitcoin), any coin Coinbase trades in US dollars. Its Coil as a whole-number score out of 100, whether it is at 100, when its last Break fired, and whether it is scored this hour or sitting out.
• list_squeezing: every scored coin at or above a Coil score (default 90), highest first, at most 50.
• recent_breaks: the coins whose Break fired in the last N hours (default 24, at most 168), newest first, at most 50.
• market_funnel: how many coins are tracked, active, squeezing, at a full squeeze, and broke out in the last 24 hours. Counts only.

Coil is a squeeze meter from 0 to 100: how tight price is vs. the last 5 days; 100 = tightest. The tools return derived readings only, never candles or prices, and never the whole list of coins in one call. At most 60 requests a minute per address. Readings refresh once an hour, after the hour closes; the live charts and the method page show the same numbers and how they are made.

Not a forecast.