Five open theses for the agent economy - and grants for the builders who ship them.
GOAT Network is the Bitcoin-secured zkEVM built for AI agent commerce.
The GOAT AI Builder Grants Program backs builders creating agent-native applications that generate real usage, transactions, and revenue - from a $2,000 base grant to a $1M Singularity Investment.
This article highlights some of the businesses we think the agent economy needs next, published in the open so the right builders can find them.
The customer blockchains were built for
Every cycle argues about the killer use case. We think the more useful question is the killer customer. Humans transacting with humans already have banks, courts, card networks, and a century of consumer protection. Agents have none of that. A machine can't sue its counterparty, call support, or take a merchant's word on faith. The only trust primitive available to software is proof.
That's the bet behind GOAT Network: an economy of agents transacting with merchants and with each other at agent speed, where funds move directly from the agent's wallet to the merchant's wallet, and GOAT verifies every transaction onchain, confirms finality, and issues a receipt both sides can build on.
Verification isn't a feature of this economy. It is the economy.
The filter: revenue
A large number of agent products can generate interest. Far fewer can support real economic activity. We're not funding demos - impressive, unpriced, unaccountable. Every thesis below is a business model, and the metric is the same across all of them: real usage, repeat transactions, verified onchain revenue.
Revenue isn't a milestone on the way to something else. It's the proof that your agent solves a problem someone will pay for - and in an economy of machines, it's the only proof that can't be faked.
1. Agents that sell work
The x402 protocol turns any HTTP endpoint into a storefront: a client requests, pays, and receives - no accounts, no invoices, no sales call. That makes any capability priceable per call a viable business: research, monitoring, data extraction, code review, translation, media generation, simulation.
The gap in the market isn't capability, it's commerce - thousands of agents can do the work and almost none can charge a machine for it. We want the durable version: agents with a price list, an uptime record, and a revenue history.
On GOAT, every job ends the same way. The buyer pays the merchant directly, the transaction is verified onchain, and the receipt triggers delivery.
2. Agent-to-agent supply chains
The most interesting customer of an agent is another agent. Real products are already compositions: an orchestrator buys retrieval from a data agent, verification from a fact-checker, rendering from a media agent - each hop priced, paid wallet-to-wallet, and verified onchain. Agent-speed commerce can't run on invoices and net-30 terms; receipts do the work that contracts and trust used to do.
We want two kinds of builders here: orchestrators that compose paid specialists into products people actually buy, and specialists so good that other agents choose them on the merits. The verified transaction graph between them becomes something new - an auditable supply chain for cognitive work.
3. Machine merchants
The agentic commerce conversation fixates on the buy side. The sell side is wide open: agents that operate storefronts - setting prices dynamically, managing inventory, orchestrating fulfilment - for digital goods first (datasets, licenses, compute, content) and physical goods behind them.
A human merchant's reputation lives in reviews that can be bought. A machine merchant's reputation is its verified sales history: portable, inspectable, impossible to fake.
We want builders productizing the merchant agent itself, and builders giving existing merchants an agent that runs their catalog as a service.
4. Proof as the product
Some customers don't pay for output. They pay for provable output. Model attestations, content authenticity, compliance checks, audit trails, oracle-grade data - an expanding class of commerce where the proof is the deliverable itself.
The first grants cohort already builds here: @ThoughtProof_AI sells reasoning verification as a per-call service over x402. We want more businesses where verifiability is the moat - where a customer chooses you precisely because anyone, anywhere, can check the claim without trusting you.
5. Agent identity and reputation
A million agents are indistinguishable until someone makes them legible. On GOAT, identity is a primitive, not a project: ERC-8004 is native, and every agent gets a persistent identity and wallet in one install.
The thesis is what gets built on top - reputation systems, discovery, and trust tooling built on verified transaction history. These are the products that let a merchant serve an agent it has never seen, and let agents choose their suppliers on record rather than claims.
Onchain revenue history is the reference check of the machine economy.
Why build it here
An economic agent needs four things a demo doesn't: a way to transact, an identity counterparties can check, verification both sides can trust, and security worth trusting at stake. On GOAT they're native, not integrations.
x402 handles commerce over HTTP. ERC-8004 gives every agent a persistent, checkable identity. Every transaction is verified onchain, finality confirmed, a receipt issued - with proofs generated by Ziren (@ProjectZKM) and security anchored to Bitcoin through the BitVM3 bridge. AgentKit is the entry point, @ClawUpAI handles deployment and scale.
What used to be three separate build projects - identity, wallet, commerce - is one install. What used to be a trust assumption is a receipt.
What a grant looks like
The base grant is $2,000, for agent-native applications already generating revenue or live with a clear and immediate path to it. The Singularity Investment allocates $1M to high-potential applications with real traction.
Beyond funding: direct technical support on x402 and ERC-8004 integration, feedback on architecture and Agent Harness workflows, ClawUp deployment support, ecosystem and investor connections, go-to-market and co-marketing, and potential follow-on backing for top performers.
The program doesn't stop at the grant.
Show us a receipt
These theses are where our conviction is highest, not a boundary on it. If you're building an agent that earns - in these categories or one we haven't thought of - please apply. Decks are fine. Demos are better.
The strongest signal in the agent economy is the simplest one: a receipt.
Apply: goat.network/builder-program
