One-liner: A protocol for launching, funding, and governing AI-run companies — where every company is a token and every token is backed by real revenue.
Building a company comes down to three things: capital, people, and iteration.
AI is replacing people. Autonomous coding tools (Claude Code, Cursor, Devin) handle execution. Platforms like Polsia already run 1,000+ companies with AI doing engineering, marketing, and outreach — $1M+ ARR, zero human employees per company. AutoResearch closes the loop on R&D. OpenClaw acts as an always-on digital worker. The execution layer is solved.
AI is replacing iteration. Polsia's AI CEO wakes up nightly, evaluates company state, decides what to build, ships it, and reports back. The cycle that used to take weeks now takes hours. We're in the middle of the exponential — fully autonomous companies are months away, not years.
What AI can't replace: taste, direction, and capital formation. Humans are still better at deciding what to build and why. And the ability to coordinate multiple sources of capital toward a single goal is what turns an idea into a real company.
That's the gap. Polsia runs companies but has no capital formation layer. Virtuals launches tokens for AI agents but those agents don't run real businesses. PumpFun proved you can coordinate capital in seconds but there's nothing on the other side of the token.
We close the loop: capital formation + AI execution + on-chain revenue.
-
Launch a company in one click. You pick a vertical (micro-SaaS, API service, content business). An AI agent spins up, deploys infrastructure, and starts building. One company = one token, issued on a bonding curve.
-
Token holders govern direction. Proposals are weighted by token holdings. Community votes on what the company should build, which markets to target, which features to prioritize. Votes are time-boxed (30-minute rounds). The AI executes the winning proposal immediately.
-
AI executes autonomously. The agent builds product, ships updates, runs marketing, handles customer support, iterates based on metrics. No human employees. The full Polsia loop but token-governed.
-
Revenue flows on-chain. Every dollar the company earns is tracked on-chain. Revenue triggers automatic buyback-and-burn of the company's token. Token value is mechanically tied to business performance — not vibes, not speculation, not narrative.
-
Livestream everything. Every action the AI takes is visible in real-time. A dashboard shows: what the AI is building, what decisions it's making, revenue in/out, token metrics. Think Twitch for AI companies. Addictive to watch, impossible to fake.
-
Leaderboard. Companies compete on revenue. Top-performing AI companies get visibility, attract more capital, compound faster. Creates a flywheel: capital flows to performance, performance generates revenue, revenue increases token value, token value attracts more capital.
- Execution tools crossed the threshold. Claude Code, Cursor, and Devin can build production software autonomously. Polsia proved 1,000+ companies can run simultaneously with zero human workers.
- On-chain infrastructure is mature. Solana handles the throughput. Coinbase Agentic Wallets give AI agents their own financial identity. Circle Arc enables nanopayments for agent-to-agent transactions.
- Legal wrappers exist. Wyoming DAO LLC supports "algorithmically managed" entities where smart contracts are the governing authority. Each AI company can be a real legal entity.
- Capital formation is proven. PumpFun did $800M in revenue from token launches. The demand to fund things quickly and permissionlessly is massive and validated.
- The IMO precedent. Ankr's Neura introduced Initial Model Offerings — tokenized fractional ownership of AI models. We extend this from models to companies.
| Runs real businesses | Tokenized | Revenue on-chain | Community governed | |
|---|---|---|---|---|
| Polsia | Yes ($1M+ ARR, 1000+ cos) | No | No | No |
| Virtuals Protocol | No (agents do tasks) | Yes (11K+ agents, $35M fees) | No | Partial |
| ai16z / ElizaOS | No (VC DAO + framework) | Yes ($2B peak mcap) | No | Yes |
| PumpFun | No (pure tokens) | Yes ($800M revenue) | No | No |
| Believe | No (tokens for ideas) | Yes | No | No |
| Botto | Yes ($6M+ art revenue) | Yes | Yes | Yes |
| Us | Yes | Yes | Yes | Yes |
Botto is the closest comp — an autonomous AI artist governed by a DAO, generating real revenue distributed to token holders. But Botto is a single entity doing one thing (art). We're a protocol for launching many autonomous companies across verticals.
Polsia — The AI company operator. Solo founder, $1M+ ARR in one month, 1,000+ companies. AI handles everything: engineering, marketing, outreach, ads. $50/mo + 20% revenue share. The execution layer works. What's missing: no tokenization, no community governance, no capital formation beyond the founder's credit card. Polsia proves the AI can run the company. We add the coordination layer on top.
Virtuals Protocol — PumpFun for AI agents. 11,000+ agents launched, $35M in fees, 140,000+ token holders. Uses GAME framework for agent creation. But the agents are personalities and tools (influencers, analysts, chatbots) — not businesses generating revenue. The token value is speculative, not revenue-backed.
ai16z / ElizaOS — First AI-led DAO. Hit $2B market cap. ElizaOS is the leading open-source framework for crypto-native AI agents. auto.fun is their agent launchpad. Strong community and developer ecosystem. But ai16z is one DAO making investment decisions — not a protocol for spinning up revenue-generating companies. ElizaOS is a framework, not a product.
PumpFun — The gold standard for permissionless capital formation. $800M+ lifetime revenue. 9M+ tokens launched. Proved that people will fund anything if you remove friction. But 98.6% of tokens are rug pulls. There's nothing behind the token — no product, no revenue, no company. We take the PumpFun mechanic and put a real business behind it.
Believe (believe.app) — Attempted "PumpFun for ideas/companies." Tweet-to-token on Solana. Hit $14.2M weekly revenue at peak. Ben Pasternak (Monkey app, NUGGS) is a real builder. But BELIEVE token is down 99.6% from ATH. Failed because: no execution layer (ideas stayed ideas), no revenue binding (token value was pure speculation), supply inflation controversy destroyed trust. Cautionary tale — the demand exists but execution and accountability are non-negotiable.
Botto — Decentralized autonomous artist. Generates 70K images/week, 5K voters select winners for auction, $6M+ in revenue. 50% to voters, 50% to treasury. Sotheby's exhibition. Proof that tokenized autonomous AI revenue works. But single-entity, single-vertical. Not a protocol.
Truth Terminal / GOAT — AI agent that promoted a token to $1B market cap. Catalyzed the entire AI agent token meta. But it didn't create a business — it created a memecoin. No revenue, no product, pure narrative.
Autonolas (OLAS) — Protocol for creating and monetizing autonomous AI agents. Raised $13.8M. "Proof of Active Agent" consensus. GnosisDAO uses their agents for governance. Infrastructure play — could be complementary plumbing for us.
| Layer | Project | Relevance |
|---|---|---|
| Agent wallets | Coinbase Agentic Wallets | AI agents hold funds, transact on-chain |
| Agent identity | BNB Chain ERC-8004 | On-chain identity for AI entities |
| Payments | Circle Arc | Nanopayments for agent-to-agent transactions |
| Legal wrapper | Wyoming DAO LLC | Algorithmically managed entity structure |
| Liquidity | Meteora / PumpSwap | Post-graduation trading |
| AI execution | Claude Code / Cursor / Devin | The agents that actually build the companies |
Protocol fees:
- X% transaction fee on company token trades (during bonding curve and post-graduation)
- % of company revenue that flows through the protocol
- Company launch fees
Token economics:
- Protocol token captures value across all companies launched
- Buyback from aggregate protocol revenue
- Governance over protocol parameters
Scaling dynamics:
- More companies launched → more trading volume → more protocol fees
- Better-performing companies → more capital inflow → more launches
- Data flywheel: every company the AI runs improves execution for the next one
-
One AI company generating real, verifiable, on-chain revenue. The single most important milestone. If we can demo "watch an AI company get created and earn its first dollar," the pitch sells itself.
-
Buyback-and-burn mechanic working end-to-end. Revenue in → token bought → token burned → supply decreases → price reflects fundamentals. Has to be visible and trustworthy.
-
The governance loop produces good decisions. Token-weighted voting needs to actually improve company performance vs. pure AI autonomy. If human governance makes the company worse, the mechanic is theater.
-
The livestream is addictive. If people watch, they trade. If they trade, there's volume. If there's volume, there's revenue. The dashboard is the distribution hack.
- What verticals first? Micro-SaaS is the obvious starting point (AI can build, deploy, and market software tools). But trading strategies, content businesses, and API services are also viable. Need to pick one and nail it.
- Token = security? Buyback-and-burn is safer than direct revenue distribution (Howey test). But this needs legal counsel. Wyoming DAO LLC is the best current wrapper.
- How much human governance vs. AI autonomy? Too much human input = slow. Too little = why have the token? The ratio matters and probably varies by company stage.
- Chain? Solana is the obvious choice (speed, cost, existing memecoin/launchpad ecosystem). But Base has Coinbase Agentic Wallets and institutional credibility.
- Competitive moat? If this works, it gets forked. The moat is: (1) the AI execution engine improving with every company it runs, (2) the community and liquidity network effects, (3) the brand/trust built through transparent livestreams.