Helen is Marketing Team Lead at Motion Trade with 4+ years in Web3 and crypto marketing. Before joining Motion Trade, she built and led the marketing function at CLS Global and managed social media campaigns for a portfolio of crypto clients at Ninja Promo. She specializes in turning complex trading products into clear stories.
Market making for AI agent tokens differs from other sectors in three ways. Prices move with the sector narrative rather than with the individual project, so risk is highly correlated and cannot be diversified within the category. Floats are often small and concentrated relative to headline valuation, which means depth has to be sized conservatively. And liquidity usually starts on-chain, on Base or Solana, before any centralised listing, so the pool and the order book have to be kept aligned from the first day. The usage behind the sector is real: x402 agentic payments on Base passed 100 million cumulative transactions in 2026. But those payments settle in stablecoins, not in project tokens, so usage growth does not automatically become buy pressure. A good liquidity plan for an AI agent token is built around that fact.
AI agents are the most discussed narrative in crypto this year, and the tokens attached to them are among the hardest assets to keep in a healthy market. The sector swings on news about models, launchpads and a handful of leading tokens, and a small project's price can move more because of a competitor's announcement than because of anything it did. This article sets out what the 2026 numbers show, what makes these tokens distinctive from a liquidity point of view, and how a market making engagement should be structured for them.
The AI agent token market in 2026
The sector is sizeable but concentrated, and much smaller than its narrative. CoinMarketCap data cited by CryptoRank in July 2026 put AI agent tokens at around 2 billion dollars in market capitalisation within a broader AI crypto sector above 25 billion dollars. The same report noted that x402 agentic transactions on Base went from close to zero in mid-2025 to more than 100 million cumulative transactions in 2026, and that the rate at which testers became paying users improved fourfold in six months. On Solana, x402 had processed 35 million transactions by March 2026.
Leading platform tokens show how wide the gap between usage and price can be. Virtuals Protocol had deployed more than 17,000 agents and generated around 39.5 million dollars in protocol revenue by early 2026, while its token sat about 86 percent below its January 2025 high, according to an April 2026 sector review. The broader picture of how AI and blockchain are converging is in AI crypto in 2026.
Why usage does not equal buy pressure
This is the single most important point for founders in the sector. Agents transact overwhelmingly in stablecoins. x402 and similar standards let an agent pay per request for data, compute or an API call in USDC or another dollar token, with payment sizes often measured in cents. None of that requires the agent, or its owner, to buy the project's token.
So a project can show impressive transaction growth while its token sees no mechanical demand from that activity at all. Token demand comes only where the design explicitly routes value to it: fees paid in the token, fees used to buy it back, staking required to operate an agent, or revenue shared with holders. A liquidity plan should be honest about which of these exist. Where none do, the token trades on narrative, and the market making engagement has to be designed for a narrative asset.
What makes AI agent tokens hard to quote
High correlation within the sector
When a leading AI token falls, the rest of the category tends to follow within hours. For a desk, that means inventory risk cannot be diversified by quoting several AI tokens, and a sharp move in one name can widen spreads across all of them at once. For a project, it means the worst liquidity days are often caused by events it does not control.
Small, concentrated floats
Many agent tokens launch with a small share of supply circulating, held by a limited number of wallets, while the headline valuation is calculated on full supply. Depth that looks adequate relative to the valuation can be very thin relative to what a single large holder could sell. Depth should be sized against the realistic largest sale in the period, not against market capitalisation, using the method in how much liquidity a token needs.
On-chain first, centralised later
Agent tokens are frequently launched through on-chain launchpads on Base or Solana and trade in pools for weeks or months before a centralised listing. When the listing arrives, the pool and the order book become two markets for the same asset, and any gap between them is arbitraged by whoever is fastest. If nobody is responsible for alignment, the gap shows up as price dispersion that exchanges monitor and holders notice. The venue choice itself is discussed in which chain to launch a token on in 2026.
Exchange scrutiny on narrative listings
Exchanges list narrative tokens quickly and remove them quickly. IOSG's analysis of Binance delistings found that sixty-three percent of the perpetual contracts Binance removed in 2026 came from Binance Alpha, and that contracts with open interest below one million dollars were delisted at a rate of thirty-one percent, as published in September 2026. Many AI agent tokens get a perpetual listing before a spot one, so open interest on that contract is part of the liquidity picture. More on this in exchange volume requirements to stay listed.
| Feature of AI agent tokens | Liquidity consequence | What the engagement should include |
|---|---|---|
| Sector-wide correlation | Spreads widen across the category on one name's news | A written volatility policy: how far the spread may widen and for how long |
| Small, concentrated float | Depth is thin relative to what one holder can sell | Depth sized against the largest realistic sale, not market cap |
| On-chain launch first | Pool and book drift apart after a CEX listing | One party responsible for cross-venue alignment from day one |
| Usage settles in stablecoins | Adoption does not create token demand by itself | A realistic demand assumption in the liquidity budget |
| Perpetuals listed early | Open interest becomes a survival metric | Spot and perp books kept consistent; OI monitored |
How to structure a market making engagement for an AI agent token
Start with the inventory. Because risk is correlated and floats are small, the desk should hold less inventory relative to valuation than it would for a comparable token in a less reflexive sector, and the contract should say how inventory is managed when the whole category moves. A loan-and-option structure deserves particular care here, since a sharp narrative rally can move options deep into the money quickly. The trade-offs are in retainer versus loan model.
Then write the volatility policy. Holding a fixed spread through a sector-wide sell-off is unrealistic, and pretending otherwise produces a contract nobody can meet. A better agreement sets the normal spread and depth, the maximum spread allowed during defined volatility events, and how quickly the desk must return to normal. The broader contract structure is in what belongs in a market making agreement.
Finally, assign cross-venue alignment explicitly. Whoever quotes the centralised book should also be responsible for keeping the on-chain pool in line, or should work with whoever manages it under a clear agreement. Motion Trade's approach for this sector is described on our AI and AI agents page.
FAQ
How is market making for AI agent tokens different?
AI agent tokens move with the sector narrative, so risk is highly correlated across the category. They often have small, concentrated floats and start trading on-chain before a centralised listing. Depth has to be sized conservatively, a volatility policy has to be agreed in advance, and pool and order book prices have to be kept aligned.
How big is the AI agent token market in 2026?
CoinMarketCap data cited by CryptoRank in July 2026 put AI agent tokens at around 2 billion dollars in market capitalisation, within a broader AI crypto sector of more than 25 billion dollars.
Does x402 usage increase demand for AI agent tokens?
Not directly. x402 payments settle in stablecoins such as USDC, so transaction growth does not create buy pressure on a project token unless the token design routes value to it through fees, buybacks, staking or revenue sharing.
How many x402 transactions have there been?
A report cited by CryptoRank found that x402 agentic transactions on Base went from near zero in mid-2025 to more than 100 million cumulative transactions in 2026. Solana reported 35 million x402 transactions on its network by March 2026.
Should an AI agent token launch on-chain or on a CEX first?
Most launch on-chain first, usually on Base or Solana, because that is where the agent ecosystems and launchpads are. The important thing is to plan the transition, so the pool and the order book stay aligned once a centralised listing arrives.
What KPIs should an AI agent token market making contract include?
A normal spread ceiling, depth at defined bands around the mid price, an uptime target, a volatility policy setting how far and how long spreads may widen during sector-wide moves, cross-venue price alignment, inventory limits and clear exit terms.
