Skip to main content

Performance

<300ms
API latency

Median response time across endpoints, measured at the edge.

Seconds
Data freshness

Most streaming data is available within seconds of the source event. Seconds-fresh by default.

99.9%
Uptime target

Redundant infrastructure with automatic failover and continuous monitoring of ingest agents.

Data freshness by type

Data typeFreshnessSource
Trades and pricesWithin secondsPolymarket, Kalshi
Orderbook snapshotsWithin secondsPolymarket
Market metadataWithin secondsPolymarket, Kalshi
Social mentionsWithin secondsX, Reddit
Candlesticks (OHLCV)Per-candle closeComputed
F-Score and wallet profilesUpdated every few hoursComputed
Market 360 scoresUpdated every few hoursComputed

Rate limit resilience

You never call source platform APIs directly. Falcon maintains persistent connections to data sources, handles rate limits, retries, and reconnections behind the scenes, and serves consolidated data through its own API.

Without Falcon
  • Manage API keys per platform
  • Handle rate limits and retries yourself
  • Normalize different response formats
  • Maintain WebSocket and polling clients
  • Monitor upstream API changes and outages
With Falcon
  • One API key, one endpoint
  • No upstream rate limit concerns for your app
  • Consistent JSON across platforms
  • Seconds-fresh, always-on ingest
  • Sub-300ms response times

MCP Streaming

The Claude MCP integration uses Server-Sent Events (SSE) for live streaming. Claude can query prediction market data mid-conversation and receive structured results with sub-second latency — no polling, no delays, no stale context.