wildsurfer/cnn-fear-and-greed-parse

Go package for CNN's Fear & Greed Index, a 0–100 gauge of US stock market sentiment. It returns the current score and rating, the values for the previous close, week, month and year, about a year of daily history, and the seven component indicators (market momentum, stock price strength, stock price breadth, put/call options, market volatility, junk bond demand, safe haven demand), each with its own score, rating and history of raw values. The module also ships a CLI and an MCP server, and has no dependencies outside the Go standard library.
Install
Usage
Output:
An Indicator's History holds the raw underlying series (the S&P 500 level for momentum, the VIX level for volatility, ratios and spreads for the rest), and its Score is CNN's 0–100 normalization. To use your own http.Client (timeout, proxy), replace cnnfag.HTTPClient.
CLI
For cron jobs and shell pipelines, without writing Go:
-json prints the full result, including the daily history. -timeout changes the request timeout (default 15s).
MCP server
cnnfag mcp runs a Model Context Protocol server over stdio, so AI assistants can query the index. It exposes one tool, get_fear_and_greed, with an optional include_history argument. Configuration for MCP clients:
The client must be able to find the binary: use the full path (usually ~/go/bin/cnnfag) if your MCP client does not inherit your shell's PATH. Like the rest of the module, the server is built on the standard library only.
The server is also published to the MCP Registry as io.github.wildsurfer/cnnfag, with a container image at ghcr.io/wildsurfer/cnnfag for clients that prefer Docker over a local binary.
How it works
CNN does not offer a documented public API. This package requests the JSON endpoint that the Fear & Greed page itself uses:
The endpoint rejects requests that do not look like they come from a browser, so the package sends browser-like User-Agent and Referer headers. This is the same data source used by the known wrappers in other languages.
A scheduled CI job runs the test suite against the real endpoint once a week, so a change on CNN's side is detected within days.
Migrating from v1
v2 is a full rewrite: CNN removed the HTML page that v1 parsed, so v1 stopped working and its data model no longer matches what CNN publishes. Update the import path first:
Then adjust the calls:
| v1 | v2 |
|----|----|
| cnnfag.Parse() | cnnfag.Get(ctx), takes a context.Context |
| Result.Now.Value (int) | Result.Score (float64) |
| Result.Now.Text | Result.Rating |
| Result.PreviousClose.Value (int) | Result.PreviousClose (float64) |
| Result.OneWeekAgo.Value, .OneMonthAgo.Value, .OneYearAgo.Value | same field names, now plain float64 |
| Result.PreviousClose.Text and other past-period labels | removed; CNN's API has no labels for past periods, but every History point carries a rating |
| Result.LastUpdateDate | Result.Timestamp, now an exact time from CNN instead of a parsed guess |