AgentViz

Image of a Character in one of the scenes
image showing Customization abilities

Inspiration

Working with coding agents can feel mysterious. Much of their work happens behind a wall of terminal output, making it difficult to understand what an agent is actually doing. We wanted to make this process more transparent, engaging, and user-friendly by turning the agent's work into something you can actually see.

What It Does

AgentViz connects to a Claude Code session and transforms the agent's activity into a 2D pixel-art visualization. Instead of watching lines of text scroll by, users can see their agents working together in a visual environment. Users can also create custom agent teams and choose their own environments.

How We Built It

We built AgentViz using Claude Code with multiple sub-agents working together throughout the development process. We used dedicated reviewer and testing agents to improve the quality and reliability of the project. We also connected the Higgsfield MCP to generate the pixel-art images and environments used in the visualization.

Challenges We Ran Into

One of our biggest challenges was getting image generations to consistently match our vision. We also had to work around the long wait times that can come with coding agents, especially when multiple agents were working on different parts of the project at once.

Accomplishments We're Proud Of

Our biggest accomplishment was turning an idea into a working piece of software. While AgentViz is still an early version, we are proud to have created a more visual, engaging, and user-friendly way to interact with AI agents.

What We Learned

We learned how powerful multiple AI agents can be when they are given different roles and responsibilities. We also learned that building an AI-powered product is not just about making the technology work—the way users experience and understand that technology matters just as much.

What's Next for AgentViz

We want to make AgentViz more expressive and customizable. Our next steps include improving the generated images, adding more scenes and environments, expanding character customization, and supporting a wider range of specialized agents.