PolyTubes — Prediction Market Data for AI Agents

Polymarket and Kalshi give you raw prices. We give you the analytics layer — trader skill scores, market aggregates, cross-platform data. Connected to your agent via MCP in one command.

The problem

	Polymarket	Kalshi	Opinion.trade
Markets	244K	1.3M	On roadmap
Trades	600M+	33M	On roadmap
Data points	183M	40M	On roadmap
Market-days	556K	1.6M	On roadmap

Every prediction market builder hits the same wall. Polymarket has three separate APIs with different schemas. Kalshi tokens expire every 30 minutes. Neither platform offers historical analytics or trader-level data.

Your agent burns thousands of tokens wrestling with raw endpoints before it can answer a single question. You want to build strategies. Instead you're debugging authentication and normalizing data — the same work thousands of other builders are doing independently.

One command

"Who are the most profitable traders on Polymarket this month?"

"Find Kalshi markets about inflation with the highest volume"

"What's the 6-dim skill score for wallet 0xabc...?"

"Show me daily OHLCV for the Bitcoin market"

Your agent discovers the tools, reads the schema, and queries structured data. No integration code, no pipeline, no maintenance.

Data shopping vs. data plumbing

The raw API way

1. Connect to 5 Polymarket APIs (Gamma, CLOB, Data, Bridge, RTDS) — different schemas, different auth

2. Add Kalshi — RSA-PSS signed requests, historical/live endpoint split

3. Write parsing code for each endpoint's unique JSON format

4. Agent burns 300–500K tokens ingesting raw responses

5. AI-generated parsing code hallucinates field names — 42% of generated code contains errors

6. No way to verify accuracy. Repeat for every new question.

~500K tokens. Days of setup. Fragile.

The data shopping way

1. list_datasets — browse 16 analysis-ready tables

2. describe_dataset — check columns, types, row counts

3. query_dataset — get exactly what you need, filtered and sorted

Structured. Verified. Done.

~3K tokens. One command to set up. Reliable.

Your agent goes data shopping instead of data plumbing. Browse the catalog, pick what you need, trust what you get. You save hundreds of thousands of tokens, hours of integration work, and the data is verified against 2.3 billion source rows.

Datasets — real rows from production

Aggregated from 2.3B raw rows. Not a firehose — analysis-ready tables your agent queries with natural language.

polymarket agg_trader_skill_features 48,359 traders

6-dim skill scores + P&L. The main table for "who makes money."

{
  "wallet": "0xc2e7...63be51",
  "total_realized_pnl_from_exits": 36968016.20,
  "profitable_exit_rate": 0.66,
  "exit_profit_factor": 4.80,
  "total_trades": 963,
  "concentration_style": "diversified",
  "timing_style": "late_joiner"
}

polymarket agg_polymarket_market_daily 555,519 market-days

Daily OHLCV + volume per market. Your starting point for market activity.

{
  "title": "Bitcoin Up or Down — March 3, 5AM ET",
  "trade_date": "2026-03-03",
  "volume": 105251.62,
  "unique_traders": 12,
  "price_open": 0.49,
  "price_close": 0.01
}

polymarket feat_trader_prediction_quality 12,215 traders

Win rates, edge per share, size-weighted accuracy for traders who hold to resolution.

{
  "wallet": "0x204f...0d14",
  "prediction_win_rate": 0.51,
  "total_edge_captured_usd": 804685.52,
  "positions_held_to_resolution_count": 10867,
  "positions_won_at_resolution": 5530,
  "positions_lost_at_resolution": 5337
}

polymarket agg_polymarket_market_trader 1,721,285 rows

Per-trader per-market breakdown. Who made money on which markets.

{
  "wallet": "0x4ce6...d017",
  "condition_id": "0x90f5...2d9",
  "trade_count": 70,
  "volume": 45648633.06,
  "buy_volume": 43663911.84,
  "avg_price": 0.001
}

polymarket feat_trader_timing 38,839 traders

Early mover %, days to resolution, time-of-day patterns.

{
  "wallet": "0xe8dd...1d86",
  "avg_early_mover_pct": 0.51,
  "avg_days_to_resolution": 85.85,
  "timing_style": "mid_market",
  "total_trades": 236961,
  "unique_markets": 4323
}

kalshi wide_kalshi_markets 8,232,308 rows

Denormalized market view with event/series metadata and prices.

{
  "ticker": "EURUSDH-23FEB2812-1.05680",
  "title": "Will EUR/USD open above 1.05680 at Feb 28 12pm?",
  "status": "finalized",
  "winning_side": "yes",
  "yes_bid": 0.0,
  "spread": 0.0
}

kalshi agg_kalshi_market_daily 1,627,961 market-days

Daily OHLCV + volume per Kalshi market.

{
  "ticker": "KXPRESVISIT-26-QAT",
  "trade_date": "2025-12-30",
  "volume": 565,
  "trade_count": 1,
  "price_close": 0
}

+ 8 more datasets: agg_polymarket_market_summary, agg_polymarket_trader_daily, agg_polymarket_trader_summary, feat_trader_execution, feat_trader_risk, feat_trader_selection, feat_trader_exit_pnl, agg_kalshi_market_summary

How we compare

The closest aggregator (Dome) was acquired by Polymarket. The remaining option (Probalytics) serves raw normalized tables. Nobody else sells derived intelligence.

	Raw APIs	Dome	Probalytics	PolyTubes
Data	Raw JSON, 5+ endpoints	Normalized REST	3 raw SQL tables	16 structured datasets
Trader scoring	None	Wallet PnL	None	6-dim skill scores
MCP support	No	No	No	Native — 5 tools
Setup	Days	Hours	Hours	1 command
Price	Free (rate-limited)	Shut down	€16–45/mo	$9/mo

Get started

Your API Key

Trial expires:

Install

Or add to .mcp.json

MCP tools

list_datasets

Browse all 16 available tables

describe_dataset

Get schema, columns, and row counts

query_dataset

Query any table with filters, sorting, column selection

search_markets

Find markets by keyword across Polymarket and Kalshi

get_trader_score

Get 6-dimensional skill scores for any Polymarket wallet

Prediction market datafor AI agents