RF Signals for AI Agents: Market Technicals over MCP
Ask an AI assistant whether a stock is trending and it will usually guess from whatever it read during training. RF Signals gives it something better to work with: current end-of-day technicals, served through the Model Context Protocol (MCP). Your agent calls a tool, gets structured numbers with a date stamp, and answers from those.
Educational only, not investment advice.
The tools
| Tool | Arguments | Returns | Tier |
|---|---|---|---|
list_signal_tickers | none | Supported tickers and latest bar date | All, including free |
get_signal | ticker | Price, SMA 50/200, RSI(14), MACD, trend per timeframe, alignment verdict, exhaustion setup (9) and countdown (13) | Scout: daily only. Analyst and up: daily, weekly, monthly |
get_levels | ticker | Support and resistance, nearest above and below | Analyst and up |
compare_ratio | tickerA, tickerB | The A/B ratio with indicators and trend | Analyst and up |
scan_signals | tickers (optional, max 25) | One summary row per ticker | Analyst and up |
The tier comes from the plan on your API key: Free is Scout, Marketing Master and Promoter are Analyst, and Promoter Pro is Strategist.
Try it with no key
The public endpoint https://app.rewardsfuel.com/api/mcp/public needs no account. It offers list_signal_tickers and a daily-only get_signal, delayed by one trading day and limited to 30 signals calls a minute per IP. It also carries the free RFWIN tools draw_winners, get_draw and verify_draw.
curl -s -X POST https://app.rewardsfuel.com/api/mcp/public \
-H 'Content-Type: application/json' \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/call",
"params":{"name":"get_signal","arguments":{"ticker":"SPY"}}}'
The result is JSON with a tier, a data_as_of date, the daily indicators, an alignment block (which on the free tier says weekly and monthly need Analyst), and a disclaimer.
Connect your agent
Create an API key in the Rewards Fuel dashboard (Account, API keys and integrations). New keys include the mcp scope. Then point your MCP client at the paid endpoint. For Cursor:
{
"mcpServers": {
"rewards-fuel": {
"url": "https://app.rewardsfuel.com/api/mcp",
"headers": { "Authorization": "Bearer rf_live_YOUR_KEY" }
}
}
}
The same server also exposes your contest tools (list_contests, create_contest, export_contestants_csv and more), so one connection covers both. The paid endpoint allows 60 tool calls a minute per key.
Questions that work well
- "Is NVDA trending on the daily, weekly and monthly charts, and do they agree?" (
get_signal) - "Where are the nearest support and resistance levels for AAPL?" (
get_levels) - "Is NVDA getting stronger or weaker than QQQ?" (
compare_ratio) - "Scan my tickers and tell me which ones have all three timeframes aligned." (
scan_signals)
Keep the answers honest
Two fields matter for anyone building on RF Signals. data_as_of says which close the numbers come from; have your agent quote it. disclaimer says "Educational only, not investment advice"; have your agent pass it on. An unknown ticker returns unknown_ticker with the supported list, so agents should check coverage with list_signal_tickers instead of assuming.
Full reference: MCP documentation and the machine-readable summary at rewardsfuel.com/llms.txt. New to RF Signals? Start with what RF Signals measures.