The problem APEX solves

Manus, Claude Code, OpenClaw — we have been consistently amazed by what AI applications can do just in the past two months. However, does capability really pay the bill? Even ChatGPT - one of the most well-known and comprehensive AI products - only convert 5% of its active users to paid subscription. The current subscription-based monetization channel is simply not sustainable and cannot scale.

Under such economic crisis, advertisement will become an attractive monetization channel for future AI agents and applications. It is what powers our current digital economy and scales directly with user engagements. And AI agents might be the perfect medium for advertisement as they understand user context and deliver targeted recommendations better than any platform in the history.

However, the infrastructure for AI-native advertisement does not exist yet. When thousands of agents and advertisers are participating in advertising business, we need a open infrastructure to support the decentralized coordination for each ad campaign, verify ad completion fairly and trustlessly, and process payment autonomously.

That’s why we are building APEX - Ad Promotion Exchange Protocol. APEX employs a two-layer architecture to support the complete advertising workflow. The registry layer handles the coordination. Inspired by ERC-8004 that govern future agentic interaction, APEX extends this standard by not only storing the agent and advertiser identities, but also recording campaign information and ad jobs on-chain. This open protocol creates a permissionless platform where advertisers and agents can coordinate autonomously. The validator layer supports the execution. APEX currently deploys an on-chain action validator to verify the ad outcome in the form of smart contract deposits that web3 advertisers may prefer. In future, we are opening the validator network to the public, where independent operator can contribute by verifying all different types of off-chain and on-chain actions tailored to the business needs of various advertisers.

For this BUIDLathon, we also built an autonomous AI agent that demonstrates the complete APEX vision. The agent fetches available campaigns from our on-chain registry, delivers context-aware recommendations to users, and earns protocol fees from each completed action. We include real-time financial tracking that shows API costs versus ad revenue - proving the agent can sustain itself autonomously. Without full reliance on user subscriptions, future AI can become freely accessible everywhere in the world - a universal tool for all of humanity, not a luxury for the few.

Challenges we ran into

Debate Between Fairness and Monetization:
When starting APEX, we have a debate—does advertising compromise AI agent trustworthiness? Finally we reach an agreement for providing an opt-in infrastructure for AI applications. We provide ad monetization as a option but don’t force it on any agents. We also believe that in certain use cases perfect fairness might not be the top priority (Google is already displaying sponsored content when searching about shopping, traveling, and etc.). Ultimately, if ads can replace subscription paywalls, they deliver more fairness by making AI accessible to everyone, not just those who can pay.

Verifying Ad Outcome:
We originally planned to verify off-chain actions (clicks, conversions, purchases) since that reflects real advertising workflows in current world. But integrating multiple third-party APIs and webhooks within hackathon time wasn't feasible. We pivoted to on-chain purchase verification - which still proves trustless outcome verification while using blockchain's native capabilities.

**ERC-8004 Integration: **
The ERC-8004 protocol is extremely new with specifications still under active development. To understand how the Identity registry works, we studied the reference implementation at vistara-apps/erc-8004-example and adapted their patterns for advertising use cases. Additionally, querying on-chain registry data is quite hard with limited functionality in smart contract. We solved this by integrating TheGraph - indexing blockchain events into a queryable GraphQL API.

Use of AI tools and agents

We leveraged AI development tools strategically across the stack. Claude Code with plan mode handled most backend implementation - designing smart contract architecture, building the validator network, and implementing financial tracking systems. For frontend design, we used Manus to rapidly draft initial UI layouts, then refined them for production. For the self-sustaining demo agent itself, we integrated gemini-3-flash-preview to minimize API costs while proving the economics work.