Skip to content
MIND

Agentic Trading and Market Making for Token Projects

MIND Research· · 10 min read

Signed mandateMid price
On this page9 sections
  1. What does a market maker do?
  2. How do token projects pay market makers, and what goes wrong?
  3. Can an agent defend a price floor?
  4. How do agents manage depth, spread and inventory?
  5. How do treasury-funded buybacks work with rules?
  6. What is the line between market making and manipulation?
  7. How does MindPad run agents with guardrails?
  8. How does a customised strategy flow produce sustainable profit?
  9. What should a project ask any market maker?

Agentic trading means software agents run market-making tasks such as quoting, rebalancing and risk checks, under rules that limit what they can do. For a token project, the goal is a market with real depth, tight spreads and controlled inventory, so that one large sale does not wreck the chart. No agent can guarantee a price. What a well-built one can do is keep liquidity where trades happen and stay inside limits that holders can inspect.

This article explains what a market maker does, how agents handle depth, spread and inventory, where a "price floor" helps and fails, and where legitimate market making ends and manipulation begins. It is not financial advice.

What does a market maker do?

A market maker posts buy and sell quotes at the same time, so other traders can trade without waiting for a counterparty. Coin Bureau calls them liquidity providers that keep orders on both sides of a market. Three terms describe the work.

  • Spread. The gap between the best bid and the best ask. The market maker earns it when it fills both sides.
  • Depth. How much size sits at each price level. A narrow spread does not guarantee depth, and depth decides how far a large order moves the price.
  • Inventory. The tokens and quote assets the market maker holds. When it accumulates tokens during a decline, those holdings can lose value faster than spread income builds up.

That last point is the main risk, and it shapes everything an agent has to do. A firm can capture spread all day and still lose money overall if inventory moves against it. Coin Bureau notes that hedges bring their own costs and execution risk. Arrakis adds that new tokens often have no derivatives to hedge with, so the market maker holds the full risk.

Depth has a measurable effect on how a token trades. Arrakis reports that 2025 token launches with under $20,000 of depth within 2% on their main centralized exchange had 237% average annualized realized volatility, against 171% for those with $165,000 or more. The gain flattens above roughly $70,000. Arrakis also reports spreads of about 0.01% to 0.05% on BTC and ETH, and several percentage points on new tokens. Those figures come from centralized exchange books, not decentralized pools, but they show the same direction. Thin books move more.

How do token projects pay market makers, and what goes wrong?

Projects pay in two main ways, and the choice changes who carries the risk.

Model How it works Main risk to the project
Loan plus call option The project lends 1% to 5% of supply for 12 to 24 months, with call options at several strike prices. No upfront cash is needed. The market maker controls the inventory, and operations are largely opaque. It may sell tokens to fund quotes, which adds sell pressure.
Retainer The project supplies tokens and stablecoins and pays a fee, typically $3,000 to $7,000 per month per exchange, according to Arrakis. Needs more upfront capital, but it is more transparent and avoids the loan model's sell pressure.

Both models can create conflicts of interest. A market maker that holds a large token loan can earn more from trading that inventory than from capturing spread, and that may not match what the project wants. The Arrakis review also lists adverse selection, thin depth in passive pools, liquidity that leaves when incentives end, and the plain fact that market makers provide liquidity, not demand.

The common failure is a project that hands over tokens and assumes the chart is now managed. The agent or firm needs a clear mandate, visible parameters and a way for the project to stop it.

Can an agent defend a price floor?

An agent can support a price level for a while, and it cannot promise one. A defended floor is a budget of resting bids and added depth below the current price. Selling pressure consumes that budget. When it runs out, the price falls through the level. Arrakis lists promises of price or volume guarantees as a warning sign when you assess a market maker, and it is right.

A floor still has uses when it is described honestly.

  • It limits one side of the agent's own behavior. MindPad's design shows a sell floor, a line below which agents never sell. That keeps the vault from adding sell pressure at weak prices. It is a constraint on the agent, and it is not a guarantee to holders.
  • It deepens the book near the price. Bids at several levels make a large market sell cost more, which dampens a sudden drop.
  • It can be rule-based and visible. A floor that is a published parameter, with a stated budget, tells holders what the strategy will and will not do.

A floor that is a secret, unlimited or implied by marketing is a different thing, and it is the thing to avoid.

How do agents manage depth, spread and inventory?

Agents follow the same logic a human desk follows, with faster loops and stricter limits. The main levers are below.

Lever What the agent adjusts Why it matters
Range or band Where liquidity sits relative to price Concentrated liquidity earns fees only while price is inside its range, so ranges need monitoring
Quote width Distance between bid and ask Wider quotes protect against fast moves, and narrower quotes attract more flow
Inventory skew Shifts quotes to rebalance holdings A market maker that is long lowers its quotes to make selling more likely
Size per quote How much sits at each level Limits the damage from one large order
Safe mode Pauses or reduces activity Protects inventory when conditions are abnormal

Concentrated liquidity makes the range point concrete. Uniswap v3 and v4 let liquidity providers pick a price band. Inside the band, capital works harder, and the same depth needs much less of it. The v3 announcement says the trade-off is that liquidity outside the range earns no fees and providers must update their ranges as prices move. A position that nobody updates sits idle after a sharp move. An agent that watches the range and shifts it inside defined bounds fixes that specific weakness.

How do treasury-funded buybacks work with rules?

A buyback uses treasury funds to buy the project's own token from the market. It can add demand, and it can also waste the treasury if it runs on impulse. A rule-based design reduces that risk.

The Tokenomics.com framework describes the elements. The funding source should be an external asset such as stablecoins or protocol revenue, not the project's own token supply. Triggers should be measurable conditions that start and pause buying. Caps should set minimum and maximum spend per period and limit participation to a share of organic volume, typically 2% to 5%. Transparency should cover the allocation logic and periodic summaries of revenue used and tokens repurchased, while keeping exact order timing private to limit front-running. The framework also states the risks. Predictable buying can become the main source of demand, and thin liquidity makes execution easy to distort. Undisclosed or ad hoc buys reduce credibility, because holders cannot model them.

An agent suits this job because rules are what it runs on. The treasury sets the budget and the triggers, and the agent executes inside them. If the budget is spent, the agent stops. A buyback that cannot say when it ends or how much it will spend is a discretionary purchase, and holders should treat it that way.

What is the line between market making and manipulation?

Legitimate market making posts real quotes that real counterparties can trade against. Manipulation creates a false picture of demand or supply. Three practices cross the line.

  • Wash trading. Buying and selling with yourself or a linked party to fake volume.
  • Spoofing. Placing orders you intend to cancel before they fill, to create a false sense of supply or demand.
  • Fake volume. Any activity whose purpose is to make a token look more traded than it is, for example to win a listing on a ranking site or exchange.

These are not gray areas in practice. The founder of Gotbit, a market-making firm, was sentenced to eight months in prison on 13 June 2025, and the firm received five years of probation and agreed to cease operations. Prosecutors said Gotbit "provided market manipulation services to create artificial trading volume for multiple cryptocurrency companies" from 2018 to 2024, according to The Block.

Coin Bureau lists signs to watch for. These include high reported volume with shallow order books, liquidity on a single exchange, large orders that vanish before execution, unusual volume spikes without real participation, and unclear terms or compensation. A project that buys "volume" is buying a risk of legal trouble.

How does MindPad run agents with guardrails?

MindPad is a Solana launchpad in development. A token launches only after backers fund it, and after the venue creates the pool, agents manage liquidity under signed mandates. The design rests on four rules.

  1. Fixed strategies, not chatbots. Each agent runs a deterministic loop with explicit inputs and limits. The steps are market data, strategy, intent, risk check, simulation and execution. The same inputs produce the same decision, and every step is logged.
  2. An AI layer that only recommends. A slower AI layer reads conditions and proposes bounded parameter changes, such as a narrower range when activity is low. It cannot sign, move funds or change a mandate. Rules decide what runs.
  3. Mandates enforced onchain. Every trade or liquidity change must fit a signed mandate. The onchain program checks limits, and the result is reconciled after finality. Agents in a swarm also draw on shared onchain budgets, so no single signal can drain a vault.
  4. Published strategy versions. Parameters and version hashes are public, so anyone can check what a vault is running. Examples shown in the design include an inventory band, range liquidity and a safe mode.

The design also draws the manipulation line in plain terms. Agents do not make self-trades or run loops between themselves, and volume has to come from real counterparties. Agents may coordinate to avoid overlapping exposure, and they never create artificial volume.

For the launch side of this system, see why Pump.fun launches fail and what a funded launch changes.

How does a customised strategy flow produce sustainable profit?

Sustainable profit in market making comes from fees and spread earned on real trades, kept larger than the losses from inventory moving against you. A strategy flow built for one token tunes the levers above to that token's volume, volatility and holder base, and caps the downside with limits.

MindPad does not promise profit. Its documented revenue comes from launch and activation fees, managed-liquidity operation fees, and a protocol share of venue fees that were actually collected, with fees disclosed per launch and revenue kept separate from vault assets. A flow earns its keep when three things hold.

  • Income comes from real counterparties, not from trading against itself.
  • Inventory stays inside bands that survive a bad week.
  • The budget is finite, visible and stops when it is spent.

A strategy that needs fake volume to look profitable is not sustainable. It fails when the audience, an exchange or a regulator looks closely.

What should a project ask any market maker?

Ask for written answers to these before you hand over capital.

  1. What are the exact terms, and who holds the inventory?
  2. What parameters will the strategy use, and can I see them?
  3. What stops it, and who can stop it?
  4. How is volume produced, and from which counterparties?
  5. What does it do in a fast decline?
  6. How will you report results, and how often?

Clear answers will reduce most of the risk described in this article. Vague answers are the warning sign.

Frequently asked questions

What is agentic trading?

Agentic trading uses software agents to run trading tasks such as quoting, rebalancing and risk checks. Safer designs keep each agent on a fixed strategy with explicit limits, log every step, and let any AI layer only suggest changes. The agent does not get open-ended control of funds. Rules and limits decide what executes.

Can a market maker guarantee a price floor?

No. A market maker can place bids and add depth below the current price, but that capital is finite. If selling exceeds the budget, the price falls through it. Arrakis lists promises of price or volume guarantees as a warning sign. Treat a floor as a risk limit on a strategy, not a promise to holders.

What is the difference between market making and wash trading?

Market making posts real quotes that other traders can trade against, and it earns from the spread. Wash trading is buying and selling with yourself or a linked party to fake volume. It is illegal manipulation in many jurisdictions. Gotbit's founder was sentenced to eight months in June 2025 for creating artificial trading volume.

How do agents help a new token's liquidity?

Agents can place and adjust liquidity continuously, keep inventory within set bands, and widen or narrow ranges as activity changes. Concentrated liquidity needs this kind of monitoring because positions earn nothing once price leaves their range. Agents help only if their limits are real, their parameters are visible and their volume comes from real counterparties.

Sources

  1. 1.What Are Market Makers in Crypto?, Coin Bureau, 2026-10-15
  2. 2.Market Makers 101, Arrakis Finance
  3. 3.Liquidity Overview, Uniswap
  4. 4.Introducing Uniswap v3, Uniswap Labs, 2021-03-23
  5. 5.Gotbit founder Aleksei Andriunin sentenced to 8 months following wash trading scheme, The Block, 2025-06-13
  6. 6.Token Buybacks: How to Design a Rule-Based Buyback Framework, Tokenomics.com, 2026-06-22

Nothing on this site is financial advice.

Share

Get MIND updates in your inbox

One email when we ship. No spam, unsubscribe anytime.