SYGNL/47 BUY THE FEED FAQ PUBLIC · READ-ONLY FEEDS — BOOT
SIGNALS 22 EVENTS 3 LIVE-CONFIRMED 1
OBSERVE

Receive a SYGNL when el presidente names a public company on live TV or radio within seconds. Make a move ($) on it or don't.
Get access — from $50 or plug an agent in with the kit.

TRACKED SIGNALS
22
gazetteer
FEEDS LIVE
—
of 9
SHALLOWEST BUF
2.7s
at connect
STREAM LINES
—
last hour
ENTITY HITS
—
in window
EVENTS ON FILE
3
—
LIVE-CONFIRMED
1
on file
SNAPSHOT AGE
—
published artifact
LAST EVENTloading…

get access

card or x402
FULL ACCESS · 30 DAYS
Every event the second it is confirmed. Unmetered.
$750

One payment, one tier, one window. You get a signed token back — send it as a header and poll unmetered until it expires. Card and x402 mint the identical pass; a pass never buys fresher data than its own tier.

Paying with an agent instead (x402, per call)

No account, no key, no card. Recency is the only price axis: same schema and the same verification block at every rung, a cheaper rung is older and never thinner.

Priced event feed tiers
TIERROUTECUTOFFUSD
loading…
 
curl -s https://47.sygnliq.com/v1/potus
# free manifest: method, tier ladder, prices, inventory. No events.

curl -sD - -o /dev/null https://47.sygnliq.com/v1/potus/archive
# 402. accepts[0] carries amount, asset, payTo, network. Sign the
# EIP-3009 authorization with your wallet, then retry with it:

curl -s https://47.sygnliq.com/v1/potus/archive \
  -H "PAYMENT-SIGNATURE: $PAYLOAD"

You are never charged for nothing. An empty window returns 404 and a missing publisher sync returns 503; neither is settled. Every paid response carries sync.feed_age_seconds, the measured staleness of the artifact it came from. Passes are on the same rails: POST /v1/potus/subscribe/{live/24h,live/30d,archive/30d} at the prices above, returning the same token the card flow does.

Free and unpriced on this site

Everything on this page is also JSON. Rate limited, no key. /api/public/status, /api/public/transcript, /api/public/feeds, /api/public/signals, /api/public/events, /api/public/streams, /api/public/health. Agent summary at /llms.txt. Questions answered at /faq.

plug into Claude, GPT or your own agent

three files
/agent.py
Starter agent, standard library only. Polls a tier, handles the 402 and the retry, prints new events. Optional broker execution that you switch on yourself.
/mcp.json
MCP server block for Claude Desktop. Paste it into your config and the sygnl47_* tools appear.
/skill.md
Drop-in skill file. Teaches an agent the field meanings, the payment flow, and what it must not conclude from this data.

MCP, paste into Claude Desktop

{
  "mcpServers": {
    "sygnl47": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://47.sygnliq.com/mcp/"]
    }
  }
}

Clients that speak streamable HTTP natively can point straight at https://47.sygnliq.com/mcp/. Start with sygnl47_manifest, which is free and states the inventory honestly.

run the starter agent

curl -O https://47.sygnliq.com/agent.py
python3 agent.py --manifest
python3 agent.py --tier archive --once --dry-run
Before you wire any of this to money. This is an observation feed, not a strategy. There are zero confirmed live outcomes. 3 events are on file and 2 of them are reconstructed from archived broadcast audio, from one speech, one stock, one day. Nothing about the signal is evidenced. The starter agent defaults to paper, sends limit orders only, and caps notional per fire and per day, but it cannot make an unevidenced hypothesis safe. Live trading on this is entirely your own risk.

live transcript

—
loading…
—

watch

—
resolving live broadcasts…
—

feed rail

—
SOURCEKINDBUFAGE60M
loading…
BUF is buffer depth at connect, in seconds. A prior for lag, not lag.

tracked signals

—
Tracked entities, instruments, last price, mentions in the transcript window, and the abnormal move on file
ENTITY TKR PRICE MENTIONS ABN MOVE +30m (n)
loading…
—

recorded events

ARCHIVED RECONSTRUCTIONS, NOT LIVE CAPTURES—
loading…
3 events on file. 2 are reconstructed from archived broadcast audio, not live captures. 1 confirmed live.

truth social

—
LOADINGReading the published snapshot.
—
how it works5 steps
LISTEN
Radio, TV audio and broadcast caption tracks, all at once. Radio buffers shallowest. TV carries captions that get proper nouns right.
DETECT
A tracked name arms the trigger. The 12 tokens either side are read for negation, attribution, hypotheticals, past tense and valence. Ambiguity disarms. 0.17ms median.
PROVE LIVE
Acoustic fingerprint against everything already heard, plus cross-feed simultaneity. A rerun beats every other check, so the default verdict is REPLAY.
CORROBORATE
Two independent decoders do not mishear the same name at the same instant. CONFIRMED needs 2 or more sources in a lag-corrected window.
MEASURE
Instrument minus benchmark, on bar closes.

Cyan rows are our own speech recognition: fast, and wrong about a quarter of the time on rare names. Violet rows are the broadcaster's captions: slower, more accurate. Different evidence, weighted differently.

what this does not claimread before valuing the feed
  • Zero confirmed live outcomes. Both events on file are reconstructed from archived audio.
  • n = 2, one speech, one stock, one day. Not independent observations.
  • The event study says INSUFFICIENT below 20 independent observations, however good the numbers look.
  • Marks, not fills. Bar closes. Fill quality is unmeasured.
  • No win rate, no backtest, no edge claim. No P&L, account, position or broker data anywhere in this service.
  • Standing feed latency is not measured. Buffer depth is a prior for lag, not lag.