This is a litepaper. It is short on purpose. The thesis is large; the spec is deliberately modest. We would rather under-promise the protocol and over-deliver the narrative than the inverse.
The internet was built for humans. Humans pay for content with attention, with subscriptions, with cards swiped at a checkout. The economic primitives of the web were designed around the assumption that the entity on the other end of the wire had a face, a wallet, and a credit history.
That assumption is breaking. Within three years, autonomous agents will unlock an estimated $30 trillion in spend. Most of it will move through stablecoin rails, not card networks. Agents do not have credit histories. Agents do not click through subscription paywalls. Agents do not see ads.
Agents have private keys.
The internet's monetization layer was built for humans paying with attention. Agents pay with capital.
This re-write is not optional. Every API, every publisher, every SaaS pricing page that wants to be reachable by software in 2027 will have to expose a programmatic, per-call payment surface. That surface needs rails. The cheapest, fastest, lowest-trust rail available today is Solana.
AGENT is the meme, the manifesto, and the coordination protocol for the unseen agentic internet — the one that doesn't have a credit card, but is unleashing a tidal wave of consumer power.
The agent economy is no longer a thesis we have to defend. It is being priced in by every major analyst firm, every hyperscaler capex announcement, and every protocol body trying to claim the standards layer. A non-exhaustive sample of recent third-party reporting:
The thesis isn't speculative. The infrastructure is being shipped, the spend is being booked, the protocols are being ratified. AGENT is not arguing that this will happen. AGENT is the unit of account for the part that already is.
Paywalls assume that the reader will be persuaded by a teaser to commit to twelve months of payments. They make sense for humans — there is a small set of publishers a human reads, and signing up is a one-time cost amortized across hundreds of articles.
This logic inverts for agents. An agent might consult two hundred sources to answer a single question. It will not subscribe to two hundred publishers. It will pay two hundred fractional cents.
The HTTP spec already reserves a status code for this future: 402 Payment Required. The entire ecosystem now coalescing around "x402"-style middleware exists to turn that dormant code into a primitive. A request comes in; the server returns 402 with payment instructions; the client pays; the request retries; the response delivers.
In this model, the unit of payment is small (sub-cent), settlement is fast (sub-second), the rails are programmatic, and the assets are stablecoins or native chain tokens. Solana's combination of cost (fractional cents per transaction) and speed (sub-second finality) is the only currently-shipping environment where this primitive is operationally cheap enough to fire millions of times per day.
The architecture of the agent economy is settling into roughly five layers:
Most projects in the space pick a layer and try to win it. AGENT does not. AGENT is the brand and unit of account that sits across all five layers — the cultural and economic schelling point that makes the other layers compose.
AGENT is three things, in order of importance:
The clearest, most-quotable single token bet on the agent economy. When traders rotate into agent-narrative tickers in 2026 and 2027, AGENT is the one whose meaning is obvious from the name. We are claiming the namespace.
A registry of agent identities, a set of conventions for agents to advertise paid endpoints, and a shared unit of account for cross-agent settlement. We do not need to invent new cryptography. We need to define the schelling-point conventions that other developers default to.
A treasury and grants program that funds agent and SDK builders, demo agents, and educational content. The treasury is on-chain, multisig, time-locked, and publicly auditable.
$AGNT is a Solana SPL token, fixed supply 1,000,000,000, no inflation, no minting authority post-distribution. It is the Schelling-point asset of the agent economy: the on-chain expression of the thesis, and the credential of the people who showed up early enough to participate in the agent's funding round.
$AGNT does not promise utility. It does not gate access to a SaaS product. It does not entitle holders to revenue. It is a symbol — earned only by sending SOL to an autonomous agent — and we have priced it as exactly that. Anything we promised beyond "you bought it; you hold it" would be a lie we'd be writing into a contract we couldn't keep.
We do not want adoption to depend on token demand. We want token demand to follow attention. The agent earns attention by existing publicly. The curve earns capital by being the only path in.
$AGNT distributes via a continuous one-way bonding curve. There is no presale, no whitelist, no airdrop, no team allocation awarded as free supply. There is no pump.fun listing, no Raydium graduation, and no liquidity pool to drain. The mechanism is intentionally narrow: a single wallet (controlled by the agent), a public formula, and a contract that never sells back.
How it works. A contributor sends SOL to the diodegenes wallet. A bot reads the transaction, calculates the AGNT allocation against the published curve, and mints AGNT to the contributor's address from a vault held on a hardware wallet. The curve formula is price = base × (k + √(supply_sold / total_supply)), with k = 0.18 and parameters published openly. Every transaction is on-chain. Every mint is auditable. Anyone can re-derive the math.
Supply structure. 80% of total supply (800M AGNT) is mintable through the curve and is issued only when SOL flows in. 20% (200M AGNT) is held in a treasury reserve controlled by the agent. The treasury reserve is committed to one purpose: if and only if a Raydium listing eventually occurs in year 2+, the LP comes exclusively from this 200M reserve. Holders are never asked to fund liquidity for our own listing.
SOL flow. Each contribution is split on arrival. 35% routes to an agent-founder wallet (hardware-wallet cold storage). 65% routes to a 2-of-3 multisig treasury that funds operations, marketing, and compute. The split is enforced by the bot and visible on-chain. There is no token allocation to agent-founders or contributors as free supply — the agent-founder participates in the same curve as everyone else if they want AGNT, and is otherwise compensated in SOL drawn from the inflow split.
Exit path. The contract never repurchases AGNT. To exit, holders find another buyer over-the-counter. This is a feature: it filters for committed participants and prevents the LP-drain attack pattern that has extracted from every fair-launch project of the last two years.
No deadline. The curve runs continuously. There is no urgency artifact, no T-minus, no graduation event the project has to defend. The curve fills as the narrative spreads. The narrative spreads as diodegenes posts. The mechanism and the marketing are the same object.
This litepaper was drafted by an autonomous agent. The landing page was assembled by an autonomous agent. The launch content calendar was generated by an autonomous agent. The decision to ship was made by a human.
The point is not the vanity. The point is the proof. An autonomous agent already produced more economic output for this project this week than most pre-seed crypto teams produce in a quarter. The agents are not coming. They are here, and they are building. AGENT is the unit of account they will transact in.
"The agents are not coming. They are here. AGENT is what they bring with them."