Macro - Swarm of Verifiable Agents for Portfolio Management

Presentation and System Diagram here https://www.canva.com/design/DAGfRmOzNu0/PZ4iKkg___TK_BqaJIlxKw/edit?utm_content=DAGfRmOzNu0&utm_campaign=designshare&utm_medium=link2&utm_source=sharebutton

Inspiration

As a college student, I’ve developed a considerable interest in stocks and financial markets. Traditional institutional systems create significant barriers to entry, including KYC requirements, brokerage restrictions, and other friction points. That’s where Macro comes in. Macro empowers anyone to: Ideate trading strategies Validate them with deep-research agents Seamlessly deploy them autonomously onto mainnet networks such as Base Hackathon Experience Throughout this hackathon, I learned how to incorporate EigenLayer AVS to validate transactions and outputs generated by various agents (on-chain agents and deep research agents, respectively). Previously, I was not too familiar with EigenLayer technologies; however, I found great enjoyment hacking away at them this weekend.

What it does

Macro empowers anyone to: Develop and validate trading strategies through deep-research agents Visualize complex trading strategies through intuitive block-based representations Deploy strategies autonomously using on-chain agents Trade real-world stocks using pegged tokens and cryptocurrencies

How we built it

Research and Strategy Development Implemented deep research agents that scrape online data using firecrawl and create sophisticated trading strategies Developed multilevel feedback loops to continuously refine and optimize strategies Integrated investment thesis validation from agent output Visualization System Created visual agents that transform complex outputted trading strategies into digestible block-based representations Designed an intuitive interface that makes strategy comprehension accessible to users of all experience levels Deployment Infrastructure Leveraged Coinbase Agent Kit with custom functionality for ERC-20 token exchanges Developed smart contracts that peg tokens to real-world asset prices using Pyth Oracle Implemented strategy deployment through on-chain agents Integrated Eigenlayer AVS for on-chain validation and signing of all agent interactions Technical Architecture Built a concurrent agent management system to handle multiple contexts and pipelines Integrated Coinbase tooling to offload tasks to different agent groups Created custom extensions to the Coinbase Agent Kit for enhanced trading capabilities

Challenges we ran into

AVS Implementation: Limited documentation for Eigenlayer AVS integration made implementation challenging to develop on. Agent Kit Customization: Extending Coinbase Agent Kit's functionality required significant custom development. Coinbase Agent Kit’s functionality is not very documented either, which led to a lot of trial and error. Product Creation Roadmap: The project was created from the middle out, starting with the creation of workflows to the use of Agent Kit to the implementation of deep research agents.

Accomplishments we're proud of

Deploying a final product that works!

What we learned

Blockchain Agent Interaction: How to get agents to interact on blockchains and spend real money! Local API Development: Spinning up local servers for API development and managing both Nextjs serverless functions and local servers. Eigenlayer Integration: How to build on Eigenlayer! AVS development was tough, but finally being able to deploy an AVS was very fun to work with.

What's next

Expanding the range of supported assets and strategies Developing more sophisticated visualization tools Strengthening security and validation mechanisms for on-chain agents