Hive

Inspiration

During our internships this past summer, we both noticed that project scheduling systems are typically very slow, out of date, and hard to use. We wanted to change that, so we made Hive, a fully automated agent-powered workflow system that allows for efficient communication of project statuses via Letta agents.

What it does

In this system, every employee is represented by a Letta agent. Projects are represented as a shared memory block, or shared context between relevant agents. Business owners can assign employees to projects, and those employee's agents will be able to read, modify, and interpret those shared memory blocks. This is much cleaner and easier to use than previous project management systems because rather than combing through massive excel files or unorganized notions, employees and admins can simply ask an agent about the status of the project, or tell it what work they've done for the project and the agent will take care of all of the organization automatically. This leads to a much more streamlined workflow, and an overall more productive system.

How we built it

First we had to get the agent generation system working. This would trigger whenever a new employee signs up because every employee needs their own agent. After this, we implemented the shared memory block flow where admins could initialize new projects and assign employees to them. Lastly, we worked on a basic frontend that is familiar and easy to use.

Challenges we ran into

The internet connection was singlehandedly the biggest limiting factor in our process. For multiple hours, we couldn't use github due to a bad connection, and even when it was working general research and progress was greatly slowed. In the beginning, we also struggled to come up with an idea, and ultimately didn't start coding until quite a few hours into the challenge.

Accomplishments that we're proud of

We are proud of delivering on our agreed upon MVP throughout the internet struggles. We also spent a lot of time learning the details of agents and the Letta flow, which resulted in a much deeper understanding than we had before.

What we learned

We learned that in the future, designing a project and setting clear roles for participants is very important. If we had prepared a little more beforehand, we would have had a lot more time, and could have gotten some more of our reach goals for this project.

What's next for Hive

We really wanted to implement MCP Server stuff into the agents, but we were ultimately unable to fully implement this due to the internet problems. In the future we hope to expand on this idea and fully implement it because it would make a huge difference in productivity overall.

Built With

• letta
• react
• supabase