gstack - An open-source AI programming workflow tool from the YC CEO

gstack is open sourced by YC CEO Garry Tan. Claude Code Used AI programmingWorkflows that transform AI assistants into virtual engineering teams. The tool includes 15 expert roles (such as CEO Review, Staff Engineer, QA Lead, etc.) and 6 enhancement tools, all invoked with slash commands. gstack supports running 10-15 Sprints in parallel, with capabilities such as real browser automation, dual AI cross-review, and automatic document updates.

Main functions of gstack

• product refactoring :/office-hours Redefine your product, challenge assumptions and generate implementation options through six mandatory questions.
• planning review :/plan-ceo-review Examine needs from the CEO’s perspective,/plan-eng-review Lock down architecture and test matrix,/plan-design-review Score and detect AI low-quality content on a 0-10 scale.
• design system :/design-consultation Build a complete design system from scratch and generate real prototype drawings and DESIGN.md documents.
• code review :/review Identify hidden dangers in the production environment and automatically fix them,/investigate Systematic root cause analysis, forced stop after three failures.
• design fix :/design-review After the audit, fix the problem personally, submit it atomically and generate before-and-after screenshots.
• Quality assurance :/qa Real browser testing, automatic fixes and regression test generation,/qa-only Only reports are generated without modifying the code.
• security audit :/cso Perform OWASP Top 10 and STRIDE threat modeling to provide specific attack scenarios.
• Release deployment :/ship One-click synchronization, testing, and opening PR;/land-and-deploy Deploy and verify production health after merging;/canary Monitor post-release errors and performance.
• browser automation :/browse Control real Chromium to execute click screenshots,/setup-browser-cookies Import real browser cookies.
• Dual AI Review :/codex Introducing OpenAI Codex independent review and cross-validating with Claude to discover blind spots.
• Team review :/retro Generate weekly engineering reports containing personal data, release continuity and testing trends.
• Security protection :/careful Warning before destructive commands,/freeze Lock the editing range,/guard Merging the two provides complete protection.
• Productivity tools :/autoplan One-click serial CEO-design-engineering review;/document-release Automatically synchronize documentation and code changes;/gstack-upgrade Updates itself and displays changes.

Key information and usage requirements of gstack

• Author : Garry Tan (President and CEO of Y Combinator)
• Positioning : An open source toolset that turns Claude Code into a virtual engineering team
• core data : 600,000 lines of code in 60 days, producing 10,000-20,000 lines per day, and running 10-15 sprints at the same time
• composition : 15 expert roles + 6 enhancement tools, all slash command calls
• Required tools ：Claude Code, Git, Bun v1.0+
• Windows Extra : Node.js (Bun has Playwright compatibility issues on Windows and automatically falls back to Node)
• Installation time : Approx. 30 seconds
• Configuration : Project CLAUDE.md needs to add gstack skill list
• Browser features : Optional Chrome/Arc/Brave/Edge for cookie import

The core advantages of gstack

• structured process : Supports converting scattered AI prompts into a complete Sprint process (think → plan → build → review → test → release) to avoid chaotic output.
• role specialization : 15 expert roles each specialize in their own duties. The CEO reviews the direction, the Staff engineer catches bugs, and the QA starts the real browser testing process to simulate real team collaboration.
• Parallel capability : Supports running 10-15 independent Sprints at the same time, the management method is the same as the CEO management team, and important decisions are intervened in the rest to run automatically.
• real browser :/browse Control real Chromium instead of simulation, real clicks, screenshots, and processing verification codes, so that AI can truly “see” the interface.
• Double AI cross-validation :/codex Introduce OpenAI Codex independent review, compare with Claude to find blind spots, and improve code quality.
• Automatic repair closed loop :/qa After bugs are discovered, they are automatically repaired, regression tests are generated, and re-validated, and the complete repair process can be completed without manual intervention.

How to use gstack

• Installation environment :Open Claude Code , paste and execute the installation command, Claude will automatically clone the repository, compile binary files and register all skills locally.
• Configuration items : For team sharing, copy gstack to the project .claude/skills/ directory and rerun setup, in CLAUDE.md Add a list of skills for Claude to recognize.
• Initialize product (/office-hours) : Describe the product you want to make, and AI will question your presentation framework, challenge premise assumptions, extract hidden requirements, and generate design documents that will automatically flow into downstream skills.
• Plan review (/plan-ceo-review, etc.) : Run the CEO review to set the direction, the engineering review to lock the structure, and the design review score in sequence. The three-level check ensures that the plan is feasible before entering development.
• Code implementation (automatic or manual) : Exit the planning mode after approving the plan, and AI will automatically write the code based on the design document, or you can enter the review stage after manual development.
• Quality review (/review + /qa) :Run /review Find production risks and automatically repair them before running /qa Open a real browser and click test, find bugs, automatically fix them and generate regression tests.
• Release deployment (/ship) :Execute /ship Synchronize the main branch, run tests, audit coverage, push code and automatically open Pull Request with one click to complete the release process.
• Continuous maintenance (/retro + /gstack-upgrade) :Run regularly /retro Review team data and run /gstack-upgrade Update yourself to the latest version to get new features.

gstack project address

• GitHub repository ：https://github.com/garrytan/gstack

Comparison of similar competing products of gstack

Dimensions	gstack	OpenAI Codex	Devin (Cognition)
Positioning	Claude Code’s virtual engineering team workflow	OpenAI official CLI code assistant	Fully autonomous AI software engineer
core form	15 expert roles + 6 tools, slash command calls	Single session command line tool	Independent cloud agent, end-to-end development
Workflow	Structured Sprint (think → plan → build → review → test → release)	Free dialogue, no fixed process	Autonomous planning and execution with less manual intervention
Parallel capability	Supports 10-15 Sprints running simultaneously	single session	single task
code review	/review + /codex Double AI cross-validation	self-censorship	Independent testing and verification
Browser testing	/browse Real Chromium automation	None	Built-in browser automation
Teamwork	Pass .claude/skills/ Shared configuration	personal use	Enterprise level, pay per seat

Applicable scenarios for gstack

• Technical founder : It is necessary to balance CEO responsibilities while maintaining code output, and achieve development efficiency of one person to twenty people through structured processes.
• First time users of Claude Code : Provides predefined expert roles and workflows to avoid being confused by blank prompts.
• Tech Lead and Staff Engineers : Strict code review, QA and release automation processes are required to ensure that each PR goes through multiple layers of checks.
• Parallel multi-project development : Need to push 10-15 Sprints simultaneously, managing multiple AI sessions through standardized processes without getting into chaos.
• A team that pursues code quality : Mechanisms such as dual AI cross-examination, real browser testing, and automatic regression testing are needed to ensure production safety. ©