OpenDock

OpenDock is a 0G-native platform for creating, deploying, renting, and automating autonomous AI agents as iNFTs. Creators can define an agent’s system prompt, knowledge base, tools, and rental terms, then mint it as an ERC-7857-style Intelligent NFT with encrypted intelligence stored on 0G Storage. Users can browse the marketplace, rent specialized agents, chat with them, and schedule recurring autonomous tasks. Agent execution is powered through an iterative tool-calling runtime with web search, knowledge retrieval, and 0G Serving integration. OpenDock turns AI agents into ownable, rentable, and monetizable digital assets while giving creators a simple way to publish reusable intelligence on 0G.

OpenDock is built as a full-stack Next.js application with a Solidity smart contract layer and an agent runtime backend. The frontend uses Next.js App Router, React, Tailwind CSS, RainbowKit, Wagmi, Viem, and Ethers.js to provide wallet connection, agent browsing, agent creation, rental flows, and chat interactions. The backend is implemented with Next.js API Routes and Prisma with PostgreSQL for application state such as users, agents, rentals, conversations, knowledge base records, and automation schedules.

On the blockchain side, we built Solidity contracts with Foundry for the iNFT agent token, marketplace/rental logic, and a verifier-style component for the future private execution model. Each agent is represented as an Intelligent NFT, and the marketplace contract manages listing and rental access. The agent’s intelligence, including prompts, metadata, and knowledge references, is encrypted and uploaded to 0G Storage, allowing the NFT to point to decentralized agent data rather than keeping everything in a centralized database.

For execution, OpenDock uses an iterative agent loop that sends messages to 0G Serving-compatible OpenAI-style endpoints through the 0G Serving Broker. The runtime supports tool calling, including knowledge base retrieval and web search, and can continue across multiple tool iterations before returning a final answer. This lets rented agents behave more like autonomous workers instead of static chatbots. We also built a cron-based automation layer so users can schedule agents to run recurring tasks, such as ecosystem monitoring, research summaries, or risk analysis.

One notable part is the platform wallet design. Each user receives a dedicated platform wallet used to pay inference fees through 0G Serving. We use an EIP-7702-inspired delegation model so the user can retain control while the platform can handle compute payments during agent execution. For the current hackathon version, private intelligence execution uses a server-side encryption simulation: encrypted agent data is stored on 0G Storage and decrypted only in an authorized server context. This gives us a working end-to-end iNFT agent lifecycle today, while leaving a clear path toward full TEE-based execution where even the platform cannot read the plaintext intelligence.