Inspiration
Developers often spend more time managing the development process than actually writing code. A simple request can require understanding a codebase, finding the right files, implementing changes, running tests, and debugging failures.
We wanted to make that workflow more autonomous.
What We Built
AXT Workspace is an AI-powered developer workspace that turns natural-language software tasks into executable workflows. Instead of manually performing every step, users can give an AI agent a task and let it plan, work with the project, make changes, and verify the result.
How We Built It
We built AXT Workspace with a desktop interface, project management, AI agents, task orchestration, persistent memory, logging, and local AI capabilities. The Commander acts as the central coordinator, while specialized agents can execute development tasks and report their progress.
We designed the system around a simple loop:
Understand → Plan → Execute → Test → Report
What We Learned
We learned that useful AI agents need more than code generation. They need context, tools, structured workflows, and verification. Making the agent's actions visible is also important for building trust and making autonomous development understandable.
Challenges
Our biggest challenge was coordinating AI-driven tasks reliably while keeping the system predictable and easy to use. We had to think carefully about agent responsibilities, project state, execution, error handling, and how to recover when an agent encounters a problem.
AXT Workspace is our step toward making software development more autonomous, while keeping developers in control.
Log in or sign up for Devpost to join the conversation.