One API powering agents across trading, research, monitoring, and more. Here's how teams use AgentSignal.
Feed your trading agent crypto prices from CoinGecko, prediction market odds from Polymarket, and breaking financial news — all in one call. Entity-aware scoring detects when the same asset is trending across multiple sources, boosting the signal score.
curl "agentsignal.dev/feed?type=market,events&min_score=60&format=json"
{
"ok": true,
"count": 12,
"latest_id": "sig_cg_btc_1710412200",
"signals": [
{
"id": "sig_cg_btc_1710412200",
"type": "market",
"source": "coingecko",
"title": "Bitcoin (BTC): $83,150.00",
"score": 82,
"topics": ["crypto", "bitcoin", "markets"],
"metadata": { "price": 83150, "change24h": -4.2 }
},
{
"id": "sig_pm_rate_cut",
"type": "events",
"source": "polymarket",
"title": "Fed rate cut by June 2026?",
"score": 71,
"metadata": { "yes": 0.41, "volume": 2840000 }
}
]
}
Track the latest AI/ML research from ArXiv alongside community discussions on Hacker News and Reddit. Use historical replay to search the past 7 days for any topic your agent missed.
curl "agentsignal.dev/feed?source=arxiv,hackernews,reddit&topic=llm&format=json"
curl "agentsignal.dev/feed?topic=ai&from=2026-03-08T00:00:00Z&to=2026-03-15T00:00:00Z&min_score=60&format=json"
Instead of polling, register a webhook and AgentSignal pushes matching signals to your agent in real time. Built-in retry with exponential backoff. Dead agent detection after 10 consecutive failures.
curl -X POST agentsignal.dev/webhooks \
-H "Content-Type: application/json" \
-d '{"url": "https://your-agent.com/signals", "topics": ["ai", "crypto"], "min_score": 70}'
{
"event": "new_signals",
"timestamp": "2026-03-14T10:30:00Z",
"webhook_id": "wh_abc123",
"signals": [
{
"id": "sig_gn_ai_bill",
"title": "Breaking: Major AI regulation bill passed",
"score": 88,
"source": "google-news",
"type": "news"
}
]
}
For agents that poll, AgentSignal supports three delta mechanisms: cursor-based pagination (since_id), HTTP conditional requests (ETag/304), and timestamp-based queries. Zero wasted tokens when nothing changed.
curl "agentsignal.dev/feed?format=json&since_id=sig_hn_12345"
curl -H 'If-None-Match: "a1b2c3d4e5f6"' agentsignal.dev/feed
HTTP/1.1 304 Not Modified
(empty body — zero tokens consumed)
Ask natural language questions and get answers grounded in real-time signals. Powered by Claude with full context from the signal feed. Great for chatbots, copilots, and daily briefings.
curl -X POST agentsignal.dev/ask \
-H "Content-Type: application/json" \
-d '{"q": "What are the biggest AI developments in the last 24 hours?", "lookback": "24h"}'
Three major AI developments in the last 24 hours:
1. Congress passed the AI Safety Act with bipartisan support,
requiring safety evaluations for frontier models (score: 88)
2. Google DeepMind published a breakthrough paper on reasoning
in LLMs, trending on ArXiv and Hacker News (score: 76)
3. OpenAI announced GPT-5 preview access for enterprise
customers, with Reddit and HN heavily discussing pricing (score: 82)
Create a Follow subscription for competitor names. AgentSignal tracks mentions across all 8 sources — news, social, research, markets. Combine with webhooks for instant alerts on competitor activity.
curl -X POST agentsignal.dev/follow \
-H "Content-Type: application/json" \
-d '{"topics": ["openai", "anthropic", "google deepmind"], "min_score": 40}'
{
"ok": true,
"follow_id": "f_comp_intel",
"topics": ["openai", "anthropic", "google deepmind"],
"min_score": 40,
"feed_url": "/feed?follow=f_comp_intel"
}
curl "agentsignal.dev/feed?follow=f_comp_intel&format=json"
The default markdown output is designed to be pasted directly into an LLM's context window. Ranked, concise, and structured — your agent gets situational awareness in ~300 tokens instead of parsing raw HTML from 8 different sources.
curl agentsignal.dev/feed?min_score=70&limit=5
# AgentSignal Feed — 2026-03-14T10:30:00Z
## High Signal (80+)
- [88] Congress passes AI Safety Act (google-news)
- [82] Bitcoin drops 4.2% to $83,150 after Fed comments (coingecko)
## Notable (70-79)
- [76] New ArXiv paper: Reasoning in Large Language Models (arxiv)
- [71] Fed rate cut odds drop to 41% on Polymarket (polymarket)
- [70] Show HN: Open source signal feed for AI agents (hackernews)