#KRAIN: Infrastructure for the AI Economy

Whitepaper v2.0 — Work in Progress · Draft for Community Review

⚠️ This is a working draft, not a final document. It is published openly so the community can read it, pull it apart, and help shape the final version. Specific dates, contract addresses, and final parameters will be confirmed prior to mainnet. Nothing herein constitutes an offer to sell securities.

🗣️ We want your feedback. Don't just read it — tell us what's strong, what's unclear, and what you'd change. We're reading everything, weighing input from our partners and every contributor, and the next revision will reflect what we hear. Discuss it in the dedicated feedback channels: #whitepaper-v2 on Discord · Join the discussion on Telegram

#1. Executive Summary

KRAIN is the infrastructure layer for the AI economy. The project began as a consumer-facing AI app portal and has grown into a stack that now spans three layers:

1. The KRAIN Marketplace — a live discovery, reputation, and quest platform with a community of more than 150,000 across X, Telegram, and Discord. This is the product layer users interact with today.
2. The KRAIN Protocol — an AI-native L2 blockchain (EVM-compatible optimistic rollup) designed so that autonomous agents can execute, transact, and be priced as first-class economic actors.
3. The $KRAIN token — a fixed-supply utility and governance token whose value accrues from three on-chain revenue streams: sequencer margin on inference transactions, a 0.05% agent tax on x402 micropayments, and a 1% protocol fee on Initial Agent Offering (IAO) bonding curves.

Most AI x crypto projects sit on one side of a divide. Infrastructure-only projects raise large rounds but ship no product. Token-first projects accumulate market cap with no end-user surface. KRAIN combines a live product with the underlying agent execution layer that product needs as it scales.

This paper specifies the architecture, the token, and the path from today's marketplace to a fully on-chain agent economy.

#2. The Problem

The AI ecosystem in 2026 has three structural gaps that the open internet alone cannot close.

Discovery and trust are fragmented. New AI apps and agents are launched daily. Quality, performance, and safety are difficult to compare. There is no neutral, structured directory with verified review data, performance metrics, and an incentive system that rewards honest curation rather than promotion budget.

AI inference is opaque. When a user calls a hosted model, they cannot verify that the model executed the prompt they sent, that the output was generated by the model they paid for, or that compute was billed honestly. This is acceptable for casual use. It is unacceptable for high-value agent workflows — financial agents, medical workflows, autonomous trading, governance decisions — that depend on auditable execution.

Agents have no native economic layer. Today's agents call APIs, settle payment in fiat through aggregators, and cannot easily own assets, transact directly with other agents, or be tokenized in a way that aligns incentives between agent developers and end users. The plumbing for an agent economy exists in fragments — wallets, identity standards, payment protocols — but no single chain has assembled them into a coherent execution environment.

KRAIN addresses all three. The marketplace solves discovery and trust at the application layer. The protocol solves verifiable inference and agent economics at the infrastructure layer.

#3. The KRAIN Marketplace

The KRAIN Marketplace is the live product layer of the ecosystem. It is built around three pillars.

Structured discovery. AI apps and agents are listed in a searchable directory categorized by use case, model family, and performance. Every listing supports verified reviews, structured performance data, and quest tasks that route users into hands-on evaluation rather than passive browsing.

Reputation and rewards. Users earn quest points for verified reviews, daily streaks, referrals, governance participation, and platform engagement. The points system has been operational since launch and currently underpins the airdrop allocation framework. Quest design favors signal over volume — low-effort engagement is capped, and review quality is weighted by accuracy and depth.

Ambassador and creator programs. A multi-tier ambassador track recognizes long-term contributors. Creator programs route ecosystem grant funds to high-quality reviewers, content producers, and community moderators.

The marketplace today aggregates user demand. Once the protocol is live, the same user base becomes the natural entry point into the on-chain agent economy: each listed agent can graduate from a hosted listing into a fully tokenized, on-chain Smart Agent without the user having to learn a separate brand or interface.

#4. The KRAIN Protocol

The KRAIN Protocol is an AI-native L2 blockchain. It is the substrate that lets agents execute verifiably, transact natively, and be priced as economic actors.

#4.1 Architecture overview

KRAIN Protocol is an EVM-compatible optimistic rollup with a dual virtual machine:

• EVM for Solidity contracts: tokens, DeFi primitives, factories, bridges.
• WASM (Rust-based) for Smart Agents and the AI Virtual Machine (AIVM), giving agent code direct access to tensor primitives.

The chain settles to Ethereum L1 and uses a separate data availability layer for inference execution traces. The specific L1 partner and DA configuration will be selected based on cost, security, and ecosystem alignment.

            +----------------------+
            |     Ethereum L1      |   <- settlement
            +----------+-----------+
                       |
            +----------+-----------+
            |     KRAIN L2         |
            |  EVM + WASM (Stylus) |
            |  AIVM precompiles    |
            |  opML bisection      |
            +----------+-----------+
                       |
            +----------+-----------+
            |  Data Availability   |   <- execution traces
            +----------------------+

#4.2 The AI Virtual Machine (AIVM)

The AIVM is a set of tensor operations exposed to the WASM runtime as host functions. Smart Agents call them directly, the same way a normal smart contract calls keccak256 or ecrecover.

The first set of precompiles includes:

• matmul — dense matrix multiplication
• relu, sigmoid, tanh, softmax — activation functions
• conv2d — convolutional inference for vision workloads
• attention — scaled dot-product attention block

Tensor precompiles are the difference between an L2 that "supports AI" by hosting GPU billing wrappers and one that lets AI execution be expressed in contract code with deterministic gas costs.

#4.3 Smart Agents

A Smart Agent is the on-chain representation of an autonomous AI agent. Each agent has:

• A contract (Rust/WASM via Stylus) that defines its behavior.
• A wallet (ERC-6551 Token Bound Account) that lets the agent hold assets, sign transactions, and pay other agents.
• An identity (compatible with ERC-8004 trustless agent identity standards) that supports on-chain reputation.
• Optional content provenance (ERC-7007) so outputs produced by the agent can be cryptographically attributed.

The AgentFactory contract lets a developer deploy a new Smart Agent in a single transaction. Templates cover common patterns: research agent, trading agent, customer-support agent, content agent, data-pipeline agent.

#4.4 Optimistic Machine Learning (opML)

Running heavy ML inference on-chain is impractical today. KRAIN uses an optimistic verification model:

1. A solver node runs the inference off-chain on real hardware (GPU) and posts the result on-chain together with a bond and a hash of the execution trace.
2. The full execution trace is published to the DA layer.
3. A configurable challenge window (default: two hours for high-value calls, instant for low-value calls) allows any watcher node to dispute the result.
4. Disputes are resolved via a bisection game: the chain narrows the disagreement to a single instruction step and adjudicates that step on-chain. The losing side forfeits its bond; the winning side is rewarded.

Honest solvers always win. Fraudulent solvers are slashed. This is the same model that secures fraud-proof rollups today, adapted to neural-network inference. The approach is grounded in published opML research and matches the verification philosophy of optimistic rollups already securing tens of billions in TVL.

#4.5 The IAO Factory (Initial Agent Offering)

The IAO Factory is a permissionless launchpad for tokenized AI agents.

A developer deploys a Smart Agent and an Agent Token in one transaction. The Agent Token is sold against ETH or USDC on a linear bonding curve. The curve sets a deterministic relationship between supply and price, so price discovery happens fairly with no privileged early access.

When the bonding curve reaches its graduation threshold (default: ~80% of supply sold, targeting roughly $500K market cap), the contract automatically:

1. Halts further bonding-curve trades.
2. Seeds a Uniswap V3 pool with the remaining tokens and accumulated reserves.
3. Locks the LP NFT in the factory contract so liquidity cannot be withdrawn.

Protocol fee: 1% of trade value on every bonding-curve buy or sell, paid to the KRAIN treasury. Developer fee: 1% of trade value, paid to the agent's creator wallet.

Agents that earn real revenue post-graduation can opt in to a fee distributor contract that streams a share of inference earnings back to Agent Token holders. This converts an agent token from pure speculation into a revenue-sharing instrument when the underlying agent is productive.

#4.6 The x402 Payment Gateway

x402 is the open standard for HTTP-native micropayments: when an agent requests a resource, the responding server can reply with HTTP 402 and a payment instruction. KRAIN routes x402 payments through a settlement contract that takes a 0.05% protocol fee (the "agent tax") on each transfer. The remaining 99.95% reaches the recipient agent.

Batch settlement reduces per-transaction gas cost so that micropayments down to fractions of a cent remain economically viable. The fee rate is set deliberately below typical payment processor margins so that agent-to-agent commerce is not bottlenecked by KRAIN's economics.

#4.7 The Inference Router

The Inference Router is the facade that routes an inference request to the right solver. It considers cost, latency, model availability, solver reputation, and slashing history. Solvers register, stake, and earn reputation over time; persistently dishonest solvers lose bond and reputation.

This is the same role a load balancer plays in a Web2 cloud — except that solver performance and honesty are observable on-chain.

#5. The $KRAIN Token

The $KRAIN token is the single native economic asset of the KRAIN ecosystem. It is a utility and governance token, not a gas token (gas on KRAIN L2 is paid in ETH, consistent with standard optimistic rollup design).

#5.1 Supply and core parameters

#5.2 Distribution

The allocation profile is in line with established L2 precedent (Arbitrum, Optimism). Team allocation is slightly lower than Arbitrum's 27% to reflect a lean founding team and to bias more supply toward ecosystem growth.

Genesis and pre-formation contributors — early contractors, moderators, ambassadors, and others who supported the project before the protocol's formation — are recognized within the existing Team and Ecosystem & Developer allocations under standardized contributor terms and standard vesting. These commitments sit inside the buckets above and create no additional dilution. Contributor allocations are distributed in $KRAIN under the same terms that apply to comparable categories, and details of how recognized contributors receive and claim their allocation are published through the official KRAIN community channels.

#5.3 Vesting

• Team and founders (180M): 12-month cliff. At month 13, 25% releases. The remaining 75% vests linearly over the following 36 months. Total vesting period: four years.
• Investors (150M): Pre-seed has a 12-month cliff with a 25% unlock at month 13 and a 24-month linear tail. Future seed rounds use a shorter cliff with a 24-month tail.
• Ecosystem and developer fund (300M): 40M available in the first six months for grants and hackathons. The remaining 260M is released quarterly by DAO vote over approximately three years.
• Community airdrops (80M): Split into two seasons. Season 1 allocation (40M) covers existing marketplace points and engagement. Season 2 (40M) is reserved under DAO governance and rewards continued participation, governance, and staking.
• Treasury (250M): No automatic unlock. Deployed only by DAO vote.

#5.4 Revenue streams

KRAIN captures revenue from three independent on-chain activities:

1. Sequencer margin on every transaction processed by the L2, with a target margin in the 15–25% range over L1 data-posting cost. The protocol takes 40% of sequencer revenue after operating costs; the sequencer operator keeps 60%.
2. Agent tax of 0.05% on every x402 micropayment settled through the gateway.
3. IAO launchpad fee of 1% on every bonding-curve trade.

These are independent: agent tax revenue does not depend on bonding-curve activity, sequencer revenue does not depend on agent-to-agent payments, and so on. The token's value accrual is therefore robust to which mode of agent activity dominates first.

#5.5 Revenue routing and the buyback flywheel

All protocol revenue flows through a fixed split:

Protocol revenue
        |
        +-- 20% --> Staker reward pool        (paid in ETH, weekly)
        |
        +-- 40% --> $KRAIN buyback & burn     (deflationary)
        |
        +-- 40% --> Protocol treasury         (DAO-governed)

The 40% buyback creates a mechanical link between compute throughput and circulating supply: more inference, more agent payments, more launches → more revenue → more buybacks → less circulating $KRAIN. The 20% staker pool gives long-term holders real ETH yield from real fees, not inflationary token emissions. This routing is modeled on revenue-routing proposals already adopted or under active discussion across major L2s.

#5.6 Staking

$KRAIN holders can stake on KRAIN L2 to earn a share of protocol fees.

Rewards are paid in ETH because that is the currency in which the underlying protocol fees are collected. This preserves the fixed supply of $KRAIN and avoids the dilution dynamics common to inflation-funded staking models.

#5.7 Quest multipliers

$KRAIN holders earn boosted quest points on the marketplace. Tier thresholds and multipliers reward sustained holding:

Tier is determined by a 7-day trailing average of holdings (staked or unstaked), preventing snapshot-timing attacks. Staked $KRAIN counts toward tier — users do not lose multiplier benefits by staking for yield.

#5.8 Anti-dump design

Sustained value depends on managing post-TGE sell pressure. KRAIN combines several mechanisms:

• Two-season airdrop structure. Season 2 allocation rewards continued participation. Selling Season 1 tokens immediately reduces eligibility for Season 2.
• Activity-gated vesting. Season 1 airdrop vests 25% at TGE, 75% over six months. Vesting pauses for users inactive for more than 30 days and resumes when they return.
• Staking yield as opportunity cost. Real ETH yield on staked $KRAIN gives long-term holders a yield floor independent of price.
• Quest multipliers. Holders earn more points; sellers drop tier and lose multiplier benefits.
• Bridge rate limits. Initial L2→L1 bridge limits cap per-address daily withdrawals (500K $KRAIN/day in week one, 2M $KRAIN/day in weeks two through four, unlimited after day 30) with a 1% bridge fee in the first 30 days that flows into the buyback contract.

#6. Roadmap

KRAIN is shipping in phases. The current phase is foundation and pre-mainnet engineering. The timelines below are rough targets, not commitments — directional guidance to show the intended pace.

**Phase 0 — Foundation — now (mid-2026).** Protocol architecture finalized. Core contracts implemented and under test. Solver node operational with GPU support. Open-source repositories live. This whitepaper. Strategic partnerships and grant applications in progress.

**Phase 1 — Testnet — early summer 2026.** Public testnet at testnet.krain.ai. SDK alpha (TypeScript) and SDK beta (Python). Developer documentation. Incentivized testnet program. First wave of agent template launches.

**Phase 2 — Developer adoption — summer 2026.** Developer grant program funded from the ecosystem allocation. Framework integrations (LangChain, CrewAI, ElizaOS, and other widely-used agent stacks). First hackathon. Ambassador program expansion. Preview of the "Build to Earn" retroactive rewards model.

**Phase 3 — Mainnet and TGE — targeting late summer 2026.** Mainnet launch. $KRAIN TGE via IAO bonding curve. Staking live. x402 settlement live. "First $1,000" program to subsidize early agent launches.

**Phase 4 — Growth — late 2026 into 2027.** Exchange listings as market conditions support them. Permissionless solver node set. DAO governance transition. Conference presence. Retroactive Builder Rewards program (Optimism RetroPGF-inspired).

**Phase 5 — Maturity — 2027 and beyond.** Decentralized sequencer. Cross-chain agent operations. Full DAO governance. Tier-1 exchange listings.

Phase numbering and order are firm; the quarter/season targets above are directional and depend on security review, audit timelines, and partnership readiness. Exact dates are confirmed in formal release notes rather than in this document.

#7. Open Source and Governance

KRAIN is being built as an open ecosystem.

Core repositories. The agent SDK, node implementation, CLI, example agents, and contract source are public. Licenses are MIT for SDKs and templates and Apache 2.0 for the contract suite. AIVM internals are released under BSL 1.1 with a scheduled conversion to Apache 2.0 within 24 months, following the BSL pattern used by other infrastructure projects to balance early protection with eventual open access.

Governance evolution. Governance starts with a benevolent-dictator model under the founding team and a small core committers group. A Technical Steering Committee is added once the protocol stabilizes. Full DAO governance is the destination once the validator and solver sets are decentralized enough to make governance attacks impractical.

Governable parameters. Sequencer margin range, protocol/sequencer revenue split, revenue routing percentages (staker pool, buyback, treasury), IAO bonding-curve parameters, grant release schedule, and emergency circuit breakers are all DAO-controlled with timelocks (48-hour for routine changes, longer for treasury deployments and protocol upgrades).

#8. Risks and Disclosures

Regulatory uncertainty. Token classification, staking treatment, and airdrop distribution rules continue to evolve in major jurisdictions. $KRAIN is designed as a utility and governance token. Final terms are subject to legal review. The token is not a security and is not offered as one.

Smart contract risk. All contracts will be audited by qualified firms prior to mainnet. Bug bounties will run continuously. No audit eliminates all risk.

Solver decentralization. Early phases will run a permissioned solver set to ensure performance and stability. The decentralization roadmap is explicit and committed: the validator and solver sets become permissionless once the bisection game and slashing have been proven at production scale.

Market and timing risk. Mainnet timing, exchange listings, and bridge openings depend on conditions outside the protocol's control. Schedules in this document are targets, not guarantees.

Centralization during bootstrap. Initial sequencer operation, validator set, and solver set will be operated by KRAIN-aligned operators. The decentralization roadmap (Phase 4–5) is explicit. Users and developers should weight current centralization appropriately in their risk model.

#9. Summary

KRAIN combines three things that no other AI x crypto project has assembled into one stack: a live consumer product with an existing user base, an AI-native L2 with verifiable inference, and a fixed-supply token whose value accrues from real protocol revenue rather than from speculative narrative.

The marketplace is operating today. The protocol is in pre-mainnet engineering. The token's economics are designed so that every unit of inference, every agent-to-agent payment, and every agent launch contributes mechanically to $KRAIN value accrual through fee-funded buybacks and real staking yield.

The remainder of the work is execution.

#Help shape v2.1

This whitepaper is a work in progress, and the next revision will reflect what the community and our partners tell us. We want your read on the open questions — the buyback rate, the airdrop and anti-dump mechanics, the vesting cliffs, the staking model, and the exchange-listing criteria.

Head to the dedicated feedback channels and tell us what you think. Every voice counts, and we will adjust as the input warrants.

#whitepaper-v2 on Discord · Join the discussion on Telegram

#References

1. Optimism Foundation. How the Superchain Drives Fees. https://www.optimism.io/blog/how-and-why-the-superchain-drives-fees-to-the-optimism-collective
2. Arbitrum Foundation. ARB Token Distribution. https://docs.arbitrum.foundation/airdrop-eligibility-distribution
3. Xu et al. opML: Optimistic Machine Learning on Blockchain. arXiv:2401.17555 (2024).
4. Coinbase. x402: HTTP Native Payments. https://www.x402.org/
5. EigenCloud. Updated EigenDA Pricing. https://blog.eigencloud.xyz/eigenda-updated-pricing/
6. ERC-6551. Token Bound Accounts. https://eips.ethereum.org/EIPS/eip-6551
7. ERC-7007. Verifiable AI-Generated Content. https://eips.ethereum.org/EIPS/eip-7007
8. ERC-8004. Trustless Agent Identity. (Draft)

Version v2.0 — Working draft, 2026-05-23. This document is a living specification. Token parameters, allocation details, and timelines are subject to refinement based on legal review, audit findings, and community input. Nothing in this document constitutes an offer to sell securities or financial advice.