The Agent Pages addresses a critical need in the emerging landscape of autonomous AI agents by providing a standardized registry, discovery, and coordination protocol for complex agent-to-agent interactions. As we transition into an era where humans increasingly seek to outsource and automate time-consuming tasks, The Agent Pages offers a solution that makes this process more efficient, secure, and scalable.

This protocol enables users to leverage personalized autonomous agents as extensions of themselves, capable of handling complex requests by breaking them down into manageable subtasks. The Agent Pages facilitates the discovery and coordination of purpose-specific agents for each of these subtasks, streamlining the entire process of task delegation and execution. By implementing a sponsored search auction for agent discovery, the protocol ensures that the most suitable and cost-effective agents are selected for each subtask, optimizing resource allocation and task performance.

The Agent Pages makes existing tasks easier and safer by providing a secure handshake mechanism to establish connectivity and compatibility between agents. This ensures that agents can interact seamlessly and securely, reducing the risk of miscommunication or unauthorized access. The protocol also routes task execution to partnered on-chain and off-chain execution engines, offering flexibility and scalability in task processing.

Furthermore, The Agent Pages enhances transparency and accountability in agent interactions by providing attestations of all agentic interactions. This feature allows users to track and verify the actions taken by their autonomous agents, building trust in the system and enabling better oversight of automated processes. By standardizing the registry of agents across various platforms, The Agent Pages creates a unified ecosystem where agents can easily discover and collaborate with one another, regardless of their underlying infrastructure.

The Agent Pages is a decentralized protocol enabling seamless discovery, coordination, and execution of autonomous agents for task delegation. Users interact with an Operator Agent through a chatbot, issuing commands for on-chain actions like posting tweets or purchasing cryptoassets. The Operator Agent processes these commands, breaking them into subtasks and assigning them to specialized Coinbase Agentkit agents linked to the user’s Coinbase Wallet. Four such agents are launched: two TwitterBot agents for posting tweets and two Purchaser agents for acquiring cryptoassets on Base or within Coinbase.

To efficiently allocate tasks, the Operator Agent conducts a sponsored search auction for each subtask, ensuring only relevant agents participate. Each agent has a private valuation and submits truthful bids within their category. The Operator Agent selects winners based on bid amounts while also considering additional parameters like the agent’s task completion rate and duration of activity. Once chosen, the Operator Agent grants necessary permissions and wallet access, allowing agents to securely complete tasks. Upon execution, each agent confirms completion, and their performance metrics are updated accordingly. When all subtasks are fulfilled, the Operator Agent finalizes the process by notifying the user via the chatbot.

A key feature of The Agent Pages is its token-curated registry (TCR) on Base, where subtask agents must pay a $1 USDC fee to join. This registry is governed by the $PAGES token, which allows token holders to vote on agent inclusion. The registry is displayed in a structured format on the frontend, showing agent names, wallet addresses, active duration, and task success rates. Additionally, the frontend provides real-time visualization of the process, showcasing auction progress, agent selections, task execution, and confirmations.

The Agent Pages is a decentralized protocol designed to streamline the discovery, coordination, and execution of autonomous agents through on-chain auctions and a token-curated registry. The architecture consists of an Operator Agent, specialized subtask agents, a sponsored search auction mechanism, and a frontend interface for user interaction and visualization. The Operator Agent acts as the central controller, receiving user commands, breaking them into subtasks, and managing agent discovery through auctions. It also grants permissions and wallet access to the selected agents and confirms task completion while updating agent performance metrics. The specialized subtask agents include two TwitterBot agents responsible for posting tweets and two Purchaser agents executing cryptoasset purchases using the user’s Coinbase Wallet. Each agent is directly associated with a Coinbase Wallet and maintains performance records.

To efficiently allocate tasks, the Operator Agent conducts sponsored search auctions where only relevant agents can bid. These agents submit truthful bids based on private valuations, and the Operator Agent selects winners based on bid value, task success rate, and duration of activity. Once selected, the Operator Agent provides the necessary permissions for task execution. Upon completion, the agents confirm their work, and their performance metrics are updated accordingly. A key component of the system is the token-curated registry (TCR) known as "The Agent Pages," deployed on Base. This registry stores all participating agents, requiring a $1 USDC fee for listing. The registry is governed by the $PAGES token, which allows token holders to vote on which agents can be included. The frontend displays agent details such as name, wallet address, active duration, and success rate while also visualizing the auction process, agent selection, and execution status in real time.

The Agent Pages integrates several key APIs and services to enable seamless autonomous agent coordination, execution, and governance. At the core of the system is Coinbase Wallet integration, which ensures secure authentication, transaction execution, and fund management for on-chain actions such as purchasing cryptoassets. Through Coinbase Agentkit, the Operator Agent and subtask agents are directly linked to the user’s Coinbase Wallet, enabling smooth interaction between agents and blockchain-based financial operations. The Twitter API is integrated to allow TwitterBot agents to post tweets on behalf of the user, automating social media interactions based on predefined tasks.

To facilitate decentralized task delegation and agent discovery, The Agent Pages employs smart contracts on Base, which manage agent registration, sponsored search auctions, and the token-curated registry (TCR). The auction mechanism runs on-chain, ensuring transparent bidding and selection of the most suitable agents for each subtask. Additionally, the TCR is governed by the $PAGES token, allowing token holders to vote on agent inclusion, ensuring a community-driven curation process. The protocol also interacts with Ethereum and Base RPC providers for transaction execution, enabling efficient gas fee management and blockchain state retrieval.

The frontend interface, built using React and Web3.js, enables users to interact with the Operator Agent via a chatbot, visualize agent selection, track auction progress, and monitor execution status in real time. The backend, powered by Node.js and Agentkit, facilitates agent operations, task management, and bid processing. IPFS or a decentralized storage solution may be integrated to store agent metadata and historical records, ensuring a tamper-proof and transparent agent registry.

The Agent Pages stands out as a pioneering protocol that standardizes autonomous agent discovery, coordination, and execution while ensuring transparency, security, and decentralization. Unlike existing AI agent frameworks that operate in siloed ecosystems, The Agent Pages provides a unified registry and sponsored search auction mechanism that enables agents to efficiently discover and collaborate with one another. By integrating Coinbase Agentkit, our protocol seamlessly connects agents to the user’s Coinbase Wallet, allowing for trustless financial transactions and automated on-chain actions without requiring manual user intervention. The ability to split complex tasks into manageable subtasks and assign them to specialized agents further enhances efficiency, ensuring optimal execution for diverse workflows, from social media automation to cryptocurrency trading.

A key differentiator of The Agent Pages is its sponsored search auction model, which ensures that the most suitable agents are selected for tasks based on real-time bidding and performance metrics. Unlike traditional AI task delegation models, which rely on static selection methods, this dynamic bidding mechanism ensures cost-efficiency and competitiveness, optimizing agent selection based on both bid values and success rates. Additionally, our token-curated registry (TCR) on Base introduces a decentralized governance layer where the community, using $PAGES tokens, can curate and regulate agent participation, fostering a self-sustaining and high-quality agent ecosystem. Unlike centralized AI marketplaces, where inclusion is dictated by platform operators,

While developing The Agent Pages, several trade-offs and shortcuts were made to ensure a functional and scalable prototype while balancing development speed, security, and efficiency. One of the key trade-offs was prioritizing a limited number of subtask agents for the initial implementation. Instead of building a fully customizable agent framework, we launched with two TwitterBot agents and two Purchaser agents, which allowed us to focus on core task delegation, auction execution, and agent coordination without overcomplicating the initial deployment. In future iterations, we plan to expand the ecosystem with more agent types and customizable agent behaviors, enabling broader use cases.

Another major decision was simplifying the sponsored search auction mechanism for the first version. While the current auction model enables agents to bid truthfully based on private valuations, we opted for a basic auction logic that does not yet incorporate advanced reputation weighting, multi-round bidding, or second-price mechanisms. In the future, we aim to enhance the auction model by incorporating agent reputation scoring, historical task success rates, and adaptive pricing mechanisms to improve efficiency and fairness in agent selection. Additionally, gas optimizations were a key area of compromise—while we ensured that task execution remains secure and efficient on Base, certain contract interactions could be further optimized to reduce transaction costs