Garry Tan introduces GStack, an open-source framework integrated with Claude Code that leverages AI agents to collaboratively handle software development tasks—from planning and design to coding and quality assurance—thereby streamlining and scaling the engineering process. By demonstrating its capabilities through a tax app example, he highlights how GStack transforms traditional development into an efficient, AI-augmented workflow, inviting developers to harness this tool to accelerate innovation.
Garry Tan, president and CEO of Y Combinator and an experienced software engineer, introduces a transformative approach to software development called the “agent era.” Drawing from his extensive background, including co-founding Posterous and building YC’s internal platform, he emphasizes that modern AI agents can collaborate like human teams through roles, processes, and reviews. To harness this potential, he created GStack, an open-source framework that integrates AI skills into a cohesive engineering team, enabling developers to build software more efficiently and effectively.
GStack is designed to work seamlessly with Claude Code and is integrated into the Conductor platform, allowing users to interact with AI agents through various specialized skills. One highlighted skill, “office hours,” mimics the startup advisory sessions at YC, guiding users through critical questions to refine their product ideas before coding begins. Garry demonstrates this by developing a tax app that automates the retrieval of 1099 tax documents from Gmail and financial institutions, showcasing how GStack helps clarify the business model, user pain points, and technical approaches through an interactive, conversational process.
The system supports iterative design and review phases, including adversarial reviews that identify and fix potential issues in the project plan. Garry illustrates how GStack generates multiple design options for the app’s user interface, allowing users to select and refine the best fit. Beyond planning and design, GStack automates coding, testing, and quality assurance by leveraging browser automation tools like Playwright, enabling AI agents to perform complex interactions such as logging into accounts, downloading documents, and running regression tests, thereby reducing the manual QA burden on developers.
Garry shares his personal workflow, running multiple parallel GStack sessions to manage various projects and pull requests simultaneously. This high-throughput approach is supported by GStack’s robust review processes and security measures to mitigate risks like supply chain attacks. The platform’s modularity and automation empower Garry to handle hundreds of code reviews and feature developments efficiently, transforming traditional software development into a scalable, AI-augmented operation.
In conclusion, Garry invites developers to explore GStack, available on GitHub, as a powerful tool to accelerate software creation in this new era. By combining AI-driven planning, design, coding, and review, GStack dramatically lowers the barriers to building software. Garry emphasizes that with these advancements, the critical question is no longer about capability but about what meaningful projects developers will choose to build, encouraging everyone to seize this unprecedented opportunity to innovate.