Data

Prediction market data.
Cross-venue, calibrated, agent-callable.

Name: SimpleFunctions
Author: SimpleFunctions

53K normalized markets across Kalshi and Polymarket. Twelve computed indicators per market, calibration history, real-time feeds, and bulk archives — available via REST, CLI, and MCP.

Open the playground →Full documentation Economic data API

GET /api/agent/world200 · 12.4 KB

{
  "index": {
    "disagreement": 42,
    "geoRisk":      29,
    "breadth":     0.26,
    "activity":     100
  },
  "movers":     [ 8 markets ],
  "opportunities": [ 12 ],
  "stableAnchors": [ 6 ],
  "divergences":  [ 4 ]
}

53K

Latest priced markets

224

CLI commands

HTTP tools

1.7M+

Settled outcomes

Intent map

Pick the canonical data page.

The existing data pages map cleanly to the main search and agent intents: live context, bulk files, historical evaluation, indicator definitions, and reuse terms.

Live AI context

/data/real-time

Use when an agent needs the current market world model, topic feeds, MCP, or public screen/query reads.

real-time prediction market data API

Bulk files

/data/download

Use for HuggingFace datasets, JSON/JSONL exports, license attribution, and offline research loads.

download prediction market dataset

Backtests and calibration

/data/historical

Use for settled outcomes, t-24h forward prices, Brier scorecards, and rolling per-market history.

historical prediction market data

Indicator definitions

/data/indicators

Use for IY, Adj-IY, CRI, EE, LAS, RV, VR, IAR, residual VR, CVR, and regime semantics.

prediction market indicators IY CRI LAS

Reuse and attribution

/data-license

Use for CC-BY-4.0 dataset terms, citation guidance, and restrictions on API substitution.

SimpleFunctions data license

Surfaces

Access surfaces

CLI

Primary local control plane for agents, analysts, scripts, JSON output, traces, and command policy tags.

$ sf world --json

API

REST/Data API for remote services, dashboards, notebooks, and agents that cannot shell out to sf.

$ curl simplefunctions.dev/api/agent/world

Download

JSON and JSONL snapshots via HuggingFace datasets, with live API twins for the same world state, calibration, and index data.

$ huggingface.co/SimpleFunctions

Real-Time

Compressed world state (~800 tokens, refreshes every 15 min). RSS feed for changes. Streaming via CLI WebSocket.

$ sf world --watch

Historical

Settled outcomes, Brier calibration by category, SimpleFunctions Index history, and 7-day rolling price history per active market.

$ curl …/api/public/calibration

Indicators

12 quantitative indicators per market: implied yield, cliff risk, expected edge, liquidity-adjusted spread, regime, more.

$ curl …/api/public/screen?iy_min=200

Economic Data

Official macro series for CPI, unemployment, GDP, payrolls, rates, and housing with optional prediction market context.

$ curl …/api/public/query-econ?q=cpi

Quick start

Four calls to get going.

World state

Compressed snapshot (~800 tokens). Index + movers + opportunities.

curl simplefunctions.dev/api/agent/world

Screen markets

Filter active contracts by quantitative indicators.

curl "…/api/public/screen?iy_min=200&tau_max_days=30"

Economic series

Official macro series with optional market context.

curl "…/api/public/query-econ?q=cpi"

CLI

Same surfaces locally. Pipes into shell, jq, agents.

npm i -g @spfunctions/cli && sf world

Formats

Output formats

01JSONCLI default, REST API responses, and MCP adapter tool returns.

02JSONLSettled outcomes and SimpleFunctions Index history on HuggingFace.

03HF datasetsCSV-ready loading through datasets, pandas, and polars without a new API key.

04RSSReal-time change feed. World-state diffs as Atom items.

05WebSocketStreaming via CLI. Orderbook + trade tape.

FAQ

Frequently asked.

How do I access prediction market data?

Public read endpoints are rate-limited by IP without authentication; contracted tiers raise the rate limit and route to heavier model tiers. Access through REST, CLI, and MCP — all serve the same normalized world state, probability, and screener surfaces.

What formats is the data available in?

JSON via the REST API and CLI. JSON and JSONL via the public HuggingFace datasets, with CSV-ready loading through datasets, pandas, or polars. RSS for real-time change feeds. Every active market includes 12 quantitative indicators alongside price and volume.

How fresh is the data?

Pricing and orderbook state refresh every 60 seconds via the proxy layer. The world-state aggregate refreshes every 15 minutes. Historical Brier calibration is computed nightly. Daily snapshots ship by 06:00 UTC.

Try a live call.

The playground runs the same endpoints these docs describe — inspect requests, responses, and headers.

Open the playground →

API reference·Docs·Live screener·Open source

Prediction market data.Cross-venue, calibrated, agent-callable.