gstack — Garry Tan's AI-powered software factory
garrytan/gstackgstack is a comprehensive toolkit of 23 opinionated AI agents and power tools, developed by Y Combinator CEO Garry Tan, designed to enable individual developers to achieve the output of a full engineering team. It integrates with Claude Code and other AI coding agents to automate various development roles.
For technical founders, tech leads, and developers using AI agents, gstack offers a structured, opinionated workflow to dramatically accelerate software development and shipping, effectively turning AI into a virtual engineering team.
The problem it solves
Individual developers and small teams often struggle to match the shipping velocity and comprehensive coverage (design, QA, security, release) of larger engineering organizations, even with AI assistance. The challenge lies in orchestrating AI for complex, multi-faceted development tasks.
What is it?
gstack is an open-source collection of 23 specialized AI agents and 8 power tools, implemented as slash commands and Markdown-based prompts. It transforms AI coding platforms like Claude Code into a virtual team of specialists—including a CEO, Designer, Engineering Manager, QA Lead, and Security Officer—to streamline the entire software development lifecycle, from planning to deployment.
Why it's getting attention
With over 120,000 stars, gstack is highly trending due to its direct relevance to the burgeoning field of AI agents and developer productivity tools. It leverages the power of large language models to automate complex engineering workflows, a critical need for solo builders and startups aiming for rapid iteration and shipping.
Key features
- ✓23 specialized AI agents simulating roles like CEO, Designer, QA, Security Officer
- ✓Integrates with Claude Code, OpenClaw, and 10 other AI coding agents (e.g., OpenAI Codex CLI, Cursor)
- ✓Markdown-based slash commands for structured interactions
- ✓Team mode for automatic updates and shared agent configurations across repos
- ✓Native OpenClaw skills for product interrogation, strategic review, debugging, and retrospectives
- ✓Automates code review, security audits (OWASP + STRIDE), QA testing, and release processes
Best use cases
- •Accelerating product development for technical founders and CEOs
- •Providing structured, role-based prompts for first-time Claude Code users
- •Automating rigorous code review, QA, and release management for tech leads and staff engineers
- •Performing security audits and root cause debugging with AI assistance
- •Planning and implementing features end-to-end using AI agents
How to install / try
gstack provides clear installation instructions. For Claude Code, it involves a single `git clone` and `setup` command, followed by adding a gstack section to CLAUDE.md. Team mode allows for auto-updating configurations within shared repositories. Installation steps are also documented for OpenClaw and other AI coding agents, typically involving cloning the repository and running a setup script specific to the host agent.
How to use
Usage is clearly documented through a 'Quick start' guide. Users initiate workflows by describing their project, then use specific slash commands like `/office-hours` for product ideation, `/plan-ceo-review` for feature planning, `/review` for code review, and `/qa` for testing staging URLs. Native OpenClaw skills like `gstack-openclaw-office-hours` can be run directly via chat for conversational tasks.
Strengths
- ✓Potentially enables significant individual developer productivity gains, as demonstrated by the creator's experience
- ✓Transforms raw AI capabilities into structured, team-like development workflows
- ✓Supports a broad ecosystem of AI coding agents, offering flexibility
- ✓Open source and freely available under the MIT license
- ✓Automates critical, often time-consuming development tasks like QA, security, and documentation
- ✓Facilitates consistent AI agent behavior and updates across development teams
Limitations & risks
- △The substantial productivity claims are based on the creator's personal experience and may not be universally replicable.
- △Requires familiarity with AI agents, command-line interfaces, and specific AI coding platforms (primarily Claude Code).
- △Initial setup can be multi-faceted, involving configurations across different AI agent environments.
- △Reliance on external AI models means performance and cost are subject to those providers.
- △The 'opinionated' nature of the tools might not suit all development styles or project requirements.
Alternatives
Who should try it — and who should skip
gstack is ideal for technical founders and CEOs who want to accelerate product development, first-time Claude Code users seeking structured AI workflows, and tech leads or staff engineers aiming to automate and standardize review, QA, and release processes with AI agents. Those who prefer traditional, manual development methods or are not invested in AI agent tooling may find it less suitable.
Frequently asked questions
gstack is an open-source collection of 23 specialized AI agents and 8 power tools, designed by Garry Tan, to help individual developers achieve the output of a full engineering team using AI coding platforms.
gstack was created by Garry Tan, the President & CEO of Y Combinator, leveraging his two decades of product building experience.
gstack primarily integrates with Claude Code but also supports OpenClaw, OpenAI Codex CLI, OpenCode, Cursor, Factory Droid, Slate, Kiro, Hermes, and GBrain.
Yes, gstack is open source and available under the MIT License.
gstack enhances productivity by orchestrating AI agents to simulate various development roles—like CEO, Designer, QA, and Security—automating tasks from planning and design to code review, testing, and deployment.