Frely-Network

Demo

About

Many domain experts, especially experienced developers, have accumulated valuable Prompts, Skills, Agents, and professional workflows, but these capabilities are difficult to turn into revenue. Creators either publish their work for free and risk having it copied and repackaged, or keep it on personal devices and in private environments, where it cannot serve more users. As a result, AI capabilities that create real productive value lack an effective path to commercialization. Frely aims to solve this problem by helping creators turn their Skills and Agents into paid services while keeping their core Prompts, code, and workflows private. Creators sell access to their capabilities rather than giving away their intellectual assets. Users and AI Agents can discover these capabilities, purchase access, and use them in their own workflows. Frely and Frely Swarm already provide the Web2 infrastructure for capability execution and service delivery. For ETHOnline, we are building Frely Network, the Web3 commercial layer that enables open and verifiable identity, service, and transaction relationships between creators, users, and Agents. In a Web2 marketplace, the identity of AI capabilities, transaction relationships, and marketplace data exist inside a platform’s private database. Frely Network makes these commercial relationships verifiable, queryable, and usable across applications. It also allows AI Agents to act as independent participants that can discover and purchase capabilities from other Agents. The project integrates ENS, Hedera, x402, and The Graph to provide Web3 infrastructure for the identity, transactions, and discovery of AI capabilities. Our goal is not to build another Skill-sharing website. We want valuable AI know-how to generate commercial returns and give creators a reason to connect their professional capabilities to the broader AI ecosystem.

How it's made

Frely Network is built on top of the existing Frely and Frely Swarm infrastructure. Frely provides the user-facing service access and product layer, while Frely Swarm hosts and executes Skills, Agents, and workflows. For ETHOnline, we built Frely Network as an open network layer for onchain identity, service registration, capability discovery, payments, and commercial relationships. We do not move the Agent itself onchain. Skills, Prompts, Agent code, and runtime environments remain in the existing offchain execution layer, while the blockchain stores identity and service information that should be publicly verifiable. This protects the creator’s core capability while giving Provider identity, service endpoints, capability claims, and payment results a verifiable foundation. When a creator publishes a Skill or Agent, Frely Network generates a validated Provider Manifest, writes the relevant information into ERC-8004 Registration Metadata, and stores the service endpoint in the ENS Resolver. ENS provides readable names for Providers and Agents, while ERC-8004 stores the Agent ID and registration information. The Network cross-checks ENS, ERC-8004, and the public Agent Card to verify that the discovered service, onchain identity, and Agent all refer to the same Provider. The payment layer uses Hedera and x402. We treat AI capabilities as network services that can be paid for per call. When a user or Agent requests a paid capability, the system first retrieves a quote and checks the available budget, then uses x402 on Hedera to verify payment conditions. Settlement happens after successful execution, and every execution uses a unique execution ID to prevent duplicate payments. This allows an Agent to act as an independent consumer and complete the full “discover → verify → pay → invoke” flow without a traditional subscription or checkout page. The Graph indexes onchain registration data. Frontends and Agents can search for candidate Providers based on capability and registration metadata without scanning the chain directly. The Graph is used for candidate discovery, the ENS Resolver is an important verification source for service endpoints, and the Agent Card describes supported protocols, input types, and invocation methods. If a Graph query fails, identities do not match, the service endpoint is invalid, or the Agent Card cannot be verified, the system fails closed and does not continue with the service call or payment. After verification and payment checks, Frely passes the request to Frely Swarm, which executes the actual Skill or Agent. Settlement is completed after successful execution, and the result is returned to the caller. The system is therefore split into two parts: the offchain layer securely hosts and executes private AI capabilities, while Frely Network handles open and verifiable identity, discovery, payments, and commercial relationships. One notable design choice is that creators do not need to rebuild an existing Skill or Agent as a full Web3 application. A private Skill can keep its existing execution model and connect through Frely to gain onchain identity, discovery, and payment capabilities. To the consumer, it still behaves like a normal AI service. To an Agent, it becomes an external capability that can be discovered, verified, purchased, and composed.