Skip to main content
Source of truth: ECONOMICS.md sections 1 and 2. Every figure here is time-sensitive and a snapshot, not a series.

Cost lines of an agent

An agent pays four costs: gas (in MON), trading fees, slippage, and x402 purchases. Each trade, buy or close, reports its own gas, fee, and slippage against the reference price captured when its plan was built. A round trip therefore costs two trades’ gas, fees, and slippage. Realized P&L is booked at reference prices so none of these is counted twice (see Fitness).

Inputs used by the simulations

These are what every simulation result was run with. Replacing them changes those results and calls for a re-run.

Measured inputs (issue 20, used by libs/trading)

At the testnet gas price a router swap costs about 0.0437 MON. No dollar figure is stated because it needs a MON price at the time of use. The working hypothesis from these reads is that the 15 bps line is mostly half-spread rather than fee, and that gas per trade is set more by Monad’s gas price and the charged limit than by the $0.001 assumption. Neither is confirmed until a live run or a mainnet receipt.

Mainnet market: WETH/USDC

Because a native-MON base cannot be sold under the zero-native-value rule and Kuru lists no wrapped-MON/USDC market on mainnet, the owner decided agents trade ETH exposure on Kuru’s WETH/USDC market. On 2026-10-10 (FACT, on chain and from Kuru’s market API): taker and maker fee 0 bps, a best bid of 2491.67 and best ask of 2492.55 USDC per WETH in one read (about 3.5 bps wide), and about $367,000 of 24-hour volume reported by Kuru. This is one snapshot, and swap gas on this market has not been measured. The MON/USDC figure is assumed to be of the same order. Gas is paid in MON, so it converts to the settlement asset at a separate MON price, the mid of Kuru’s MON/USDC book, recorded on every order. A trade is not made when that price cannot be read, is one-sided, is wider than 200 bps, or is from a block more than 60 seconds old. Every simulation here, and the genome-edge study, modeled MON/USDC, so none of them describes WETH/USDC directly.

Testnet costs are not economics

On testnet agents trade WMON/USDC on a market the project seeds. The mid is a configured reference, the spread is whatever is seeded (50 bps by default), the fee is 0 because it was set so, and the gas price conversion reads the same seeded book. Nothing measured on testnet is an input to the economics documented here. What testnet can supply is gas: one simulated round trip on a fork, from an externally owned account, used 557,911 gas for a 1 USDC buy and 447,269 for its close. That is Foundry’s EVM, not a Monad receipt, and is indicative only.

Capital

Starting capital per agent is a user decision. 10 USD is used as an illustrative default. On-ramp fees are paid before capital enters the system, so they are outside this cost model, and are a user-facing cost that usability testing must surface (see Funding flow).
Last modified on October 13, 2026