AgentOS

LLMs are blackboxed; LLMs are uncustomizable; Introducing AgentOS: the first platform that lets you build highly customizable and explainable multi-agentic workflows for infinitely complex tasks.

Comment 1

Please watch our video

https://www.loom.com/share/58cb924eabef489a8ef5ef0b0388bbcd?sid=3ffeb455-29d1-439b-9104-b72cca99976e

Inspiration

Transform Human-Computer Interaction.

Our inspiration stemmed from the desire to create teams that seamlessly collaborate. Large Language Models are more potent than ever, able to asynchronously and autonomously run commands, peek into live data, perform extremely dense mathematical and analytical operations. They have the scope to completely revolutionize and transform the way everyone from your common Joe to executives across industries and geographies, live their lives. **

But, why haven't they?

**

To even get started with building the simplest of applications, it takes dense coding knowledge, time, resources, and more. When it comes to complex tasks that humans would love to automate, we are talking about spending hundreds of hours, and thousands of dollars to make.

AgentOS changes all of this. In less than 2 minutes, users can create infinitely complex LLM workflows and build robust and powerful teams to work on their behalf across all scopes, tasks, and complexities.

This approach allows for more dynamic and interactive workflows, reflecting real-world team interactions and processes.

What It Does

With AgentOS, users of all kinds can make extremely useful, and powerful workflow automations in seconds with one sentence. Here are some examples:

(1) Automatic Small Business Accounting Team: Seamlessly integrate with your financial books for real-time tax prep, receipt/invoice reconciliation, and expert insights from a dedicated team of AI accountants.

(2) Junior Financial Analyst Team: Research info across sites, forums, and live stock market exchange data, perform technical and qualitative analysis, create presenations, documents, and charts with advice for portfolio selection.

(3) Live Datadog Ticket/Issue Manager: Create an support-engineering team to automatically resolve On-Call issues. Monitor and identify issues in Datadog, auto-assign tickets to engineers, and resolve simple tickets

AgentOS empowers users to easily construct and manage complex workflows. Users can create specialized teams of workers, each designed to handle specific tasks, connect them to live data, knowledge bases, and live APIs, and then coordinate these teams through a supervisory agent - just like a real world Manager would. This supervisor orchestrates the workers, ensuring efficient collaboration and task completion. The result is a robust and flexible system that can adapt to various workflow needs, enhancing productivity and efficiency.

How We Built It

Our development process involved using Node.js and TypeScript for the frontend, ensuring a responsive and user-friendly interface.

We architected our own novel LLM Architecture. This is fundamental and a revolutionary approach that we aim to continue to work on in an academic setting. This includes low latency multiagent framework that expands on the capabilities of both Microsoft Autogen and Langchain as a superior way of connecting LLM Agents and orchestrating them towards complex, team-like settings.

The backend was built using Python, leveraging its powerful libraries and frameworks for seamless integration. We chose Supabase as our database solution for its real-time capabilities and ease of use. This combination of technologies allowed us to create a cohesive and highly functional web application.

Challenges We Ran Into

We ran into problems with using new libraries and languages. This is our first hackathon so understanding and building with new technologies in short crunch time pushed us out of our comfort zone. Connecting the backend logic to our frontend was extremely difficult .

Accomplishments That We're Proud Of

Firstly, we are most proud of building at our first hackathon!

We are most proud of architecting our own novel LLM architecture, that is top-of-the-line in being able to reason, deduce, and delegate tasks across teams of LLMs rapidly, and efficiently, to get resolution on large, complex, and integrated tasks.

We are proud to have developed a fully functional, end-to-end web application that meets our initial vision. Our system is not only operational but also highly efficient, showcasing the potential of combining innovative concepts with cutting-edge technology.

What We Learned

Through this project, we learned the importance of starting early and maintaining a rapid development pace. This approach enabled us to iterate quickly, incorporating feedback and improvements continuously, leading to a polished final product.

What's Next for AgentOS

Looking ahead, our primary focus is on Go-To-Market (GTM) strategies and fundraising. We are excited to bring AgentOS to a broader audience, demonstrating its capabilities and securing the necessary resources to scale and enhance our platform further. We strongly believe that our platform will revolutionize human-machine interaction, making it extremeley easy to onboard and assign complex tasks to your personal team of agents.

Built With

Share this project:

Updates