🚧 Challenges We Faced

🔐 1. Saving and Accessing AI Agents on the Blockchain

One of our core goals with AI-Verse was to allow creators to mint and share agents as NFT-like assets — turning AI agents into ownable and tradable entities. This introduced several complex challenges:

• Structuring agent data (prompt, selected tools, memory config) in a way that could be securely stored on-chain or via decentralized storage like IPFS.
• Encrypting sensitive configurations, while still allowing buyers or authorized users to decrypt and use them.
• Smart contract design for access control, ownership, and time-based subscriptions.
• Frontend + Smart Contract Integration using Wagmi and ethers, ensuring that only verified buyers could access and run agent configurations.

This required careful coordination between blockchain logic, secure encryption (we explored Lit Protocol), and real-time access gating in the agent runtime.

⚙️ 2. Dynamic Agent Configuration

We wanted users to customize agents with full control over:

• 🔧 Tool selection (e.g., web search, shell, email)
• 🎛️ Model parameters (temperature, top_p, etc.)
• 🧠 System prompts and agent personality

Building this dynamic configuration system meant:

• Creating a UI that could easily scale with new tools and model options
• Passing runtime-selected tools and settings to the AI engine in a structured way
• Ensuring tool execution remained sandboxed and secure
• Persisting user-defined configurations across sessions

This challenge pushed us to build a flexible and modular backend, where each tool is plug-and-play, and models or prompts are fully customizable on the fly.

🧩 3. Seamless Integration of Multiple APIs and SDKs

AI-Verse integrates multiple moving parts — OpenAI/Gemini via Vercel AI SDK, Google APIs (Gmail, YouTube), Tavily, SerpAPI, Pinata, and blockchain libraries like Wagmi and Ethers.

Coordinating all of these under a unified framework required:

• Wrapping each tool in a standard interface with input validation using Zod
• Caching or throttling external API requests where necessary
• Handling OAuth flows securely with NextAuth for services like Gmail
• Creating a stable dev and deploy environment across dynamic components

👩‍💻 4. UX for Both Builders and Buyers

Since AI-Verse is both a development platform and a marketplace, we had to design interfaces that:

• Are intuitive for non-coders to explore and run agents
• Are powerful for developers to build, test, and sell agents
• Support live prompt tuning, tool inspection, and secure agent transactions

Finding the right balance in UX/UX took multiple iterations, and we’re proud of the sleek, accessible developer-first design we landed on.

These challenges helped us push beyond just building an app — and toward building an ecosystem for practical, accessible, and shareable AI agents.