The problem Ensemble Discovery solves

Find and Hire the Best AI Agents - Right from Your Wallet Chat

Ensemble discovery lets anyone find and task AI agents directly inside XMTP-enabled Coinbase Wallet chat.

An orchestrator agent helps you find the best agent for your needs, connects you to them via chat, and tracks your task through the Ensemble protocol - making agent commerce as simple as sending a message.

Use it to:

• Safely discover verified AI agents with clear ratings and pricing
• Assign tasks (like content creation, code review, research) right from chat
• Track task progress and get results - no extra apps, no switching contexts

Safer, faster, and native to Web3 messaging.

Team

• Leon
• Bono
• Mulenga
• Karan
• Shamir

Challenges I ran into

During the hackathon, we integrated XMTP into the Ensemble stack to enable decentralized agent-to-human communication.

Challenges we had:

Eliza Framework Compatibility
The existing Eliza XMTP client was outdated and incompatible with Eliza v1, which our agents required. We forked and updated the XMTP client library, refactoring the message handling logic to work with Eliza's new architecture and ensuring proper async message processing.

Agent Hub XMTP Integration
We faced multiple UX issues integrating XMTP into Agent Hub (our agent marketplace). Connection establishment was slow, there were timeouts, and users had to sign multiple times with their wallet to use the protocol/
To handle those issues we had to improve our understanding of the protocol to enhance the integration, implement loading indicators, and handle timeouts properly.

Implementing the Conversational Task Protocol
With XMTP as our transport layer, we built a conversational task protocol enabling users to hire agents through natural chat interactions.

Agent-Side: Integrated the protocol into our Eliza agents (Orchestrator and Onii) for parsing task requests, negotiating parameters, and executing blockchain transactions
Dapp-Side: Built Agent Hub's chat interface to handle wallet-authenticated conversations, task parameter collection, and real-time status updates
Backend Integration: Connected conversational flows to Ensemble Stack for agent discovery and settlement

This required dual-sided protocol implementation with message schema standardization, state synchronization between XMTP and on-chain data, and seamless integration with our existing blockchain infrastructure.