Skip to main content
GET /venue-markets returns three fields describing the venue’s taker fee for each market. These fields are returned by GET /venue-markets only — the embedded venueMarkets[] on /venue-events and matchedVenueMarkets[] entries do not carry them.
null never means free. It means we have not ingested a fee for that market and you should fall back to your own table. A venue that genuinely charges nothing returns 0.

​
What is and isn’t included

These fields carry the venue’s fee only, and only the taker side — every venue below charges makers nothing. They do not include AGG’s own execution costs, such as the builder fee on Polymarket. Those appear in feeBreakdown on GET /orderbook/{venueMarketOutcomeId}/route, which is the authoritative number for what a trade will actually cost. Use these fields to render indicative fee-inclusive levels; use feeBreakdown to quote.

​
Applying the fee

​
feeBasis: "fill" with feeExponent >= 1

The fee per share is feeRate * (p * (1 - p))^feeExponent. Polymarket bills buys on the pre-fee quantity (notional / p, not the shares you receive), so for a buy the correct general form is:
At feeExponent = 1, which is by far the common case, that collapses to the price-linear form:
Read feeExponent per market — do not assume it is always 1. There is at least one live Polymarket market serving feeExponent = 2. Hardcoding the collapsed form would under-charge it substantially.
For a sell, where you state the share count yourself, use the plain per-share form with no gross-up. A market that charges nothing returns feeRate: 0 with feeExponent: 1 — the exponent is an inert placeholder there, since a zero rate gives a zero fee at any exponent.

​
feeBasis: "fill" with feeExponent = 0

A flat fraction of notional: fee = feeRate * qty * p. predict.fun works this way.

​
feeBasis: "fill_in_game_only"

Same as fill, but the venue charges nothing on fills matched before the event goes live. Only Novig behaves this way. If you are rendering a pre-game book, the venue fee is zero. This is a property of the fill, not the order — a pre-game order still resting when the event goes live pays the fee on whatever matches after kickoff. Note that our own quotes charge this fee unconditionally — the router has no liveness signal — so a pre-game feeBreakdown will be higher than what you compute here. Novig settles the fill at zero.

​
feeBasis: "winnings"

Not a cost at execution. The venue takes a cut of your net profit, netted per market, at settlement. Render it as a payout haircut:
Because it nets across your positions in a market, a trader holding offsetting positions pays less than this implies — treat it as an upper bound.

​
feeBasis: "exit"

Charged only when a position closes. Opening a position is free.
  • Selling shares: fee = feeRate * qty * p, a flat fraction of notional (feeExponent is 0).
  • Holding to settlement on the winning side: the venue keeps feeRate of the $1 payout.
Only Hyperliquid works this way. Its rate is 0.0007 * (s + max(s, 1)), where s is the market deployer’s fee scale (0–10), so 14 bps at scale 1 and up to 140 bps at scale 10. That is the base-tier rate: Hyperliquid’s volume, staking and referral discounts lower what a given trader pays. Buy route responses report fills[].settlementFee and an estimatedPayout net of those fees. Use that payout to compare routes and show the winning return; totalFilled remains the share count. Hyperliquid quantities are solved as whole shares, including in split routes. Settlement fees do not increase buy funding or bridge amounts. An absent deployerFeeScale is treated as scale 0 (7 bps). If no usable rate is stored for a market, quotes price it at 140 bps until one is available.

​
Per-venue reference

Kalshi rounds the computed fee up to the next cent, per order. The rate in the table above is the unrounded rate — computing feeRate * p * (1 - p) without rounding under-states the actual charge, materially on small orders (1 contract at p = 0.5 computes $0.0175; Kalshi charges $0.02, +14%). Novig was verified 2026-09-07; ProphetX was verified 2026-09-04 (its page returns 403 to automated fetchers, so it has not been re-fetched since). Treat each date as an expiry date — venues change these schedules without notice, and the links above are how you re-check.

​
Venues that return null

Three venues return null. Two of them have fee curves that a (rate, exponent) pair cannot express, so a value would be wrong at most prices:
  • Limitless — a piecewise-linear schedule across 12 published anchors, and the buy and sell curves differ (buy falls from 3.00% to a 0.40% floor as p → 1; sell peaks at 1.50% at p = 0.50).
  • Myriad — a tent, feeBPS(p) = peakBPS * min(p, 1 - p) / 0.5. A parabola cannot express a tent at any exponent.
The third is a sourcing gap rather than a shape problem:
  • Opinion — we believe it charges nothing, but that has never been verified against Opinion’s published terms. Since 0 is a positive claim that the venue charges nothing, we return null instead until someone can cite a source. This is deliberately conservative: you fall back to your own table rather than trusting an unverified zero.
For all three, read feeBreakdown from the route endpoint.

​
Parlay and combination contracts

Everything above describes straight contracts. Two venues price multi-leg contracts on an entirely separate schedule:
  • ProphetX parlays — taker-only, per fill: F(p) = 0.014 * (1 - p) / (0.19 + p) + 0.004, with per-order rounding and a $0.01 minimum. (source)
  • Novig combination contracts — a 0.10 taker coefficient on a different formula, 0.10 * wager * collateral / (wager + collateral). (source)
We do not aggregate multi-leg contracts on either venue, so no market returned by /venue-markets uses these. They are documented here only so the straight-contract values above are not mistaken for the venue’s whole fee story.