Key takeaways
- A free, open-source toolkit turns Claude Code into a virtual team: a CEO, a designer, a reviewer and a QA lead, all as slash commands.
- Plans get challenged before code is written: the office hours step pushes back on your framing and writes a design doc the other skills read.
- Review hunts for bugs that pass CI but break in production, and QA opens a real browser to reproduce them.
- It started on Claude Code, and other agents are supported too, but those installs are still labelled experimental.
- The catches: Windows support is minimal, outside reviews are on by default, and we have not tested any of it.
What it is
gstack turns Claude Code and other coding agents into a virtual engineering team, built from slash commands and Markdown. It is published by Garry Tan of Y Combinator under the MIT licence, and the code is on GitHub at garrytan/gstack.
The problem it solves
A blank prompt gives an agent no structure: nobody questions the plan, reviews the code or checks the result. The README pitches gstack at first-time Claude Code users who want structured roles instead, and describes it as the software factory its author uses every day.
What you get
- A sounding board: the office hours step pushes back on your framing and writes a design doc the other skills read.
- Fewer surprises: review finds bugs that pass CI but break in production, and fixes the obvious ones.
- Checks in a real browser: QA opens one, reproduces bugs and fixes them.
- A route to release: the ship step runs tests, review and the pull request.
- One joined-up process: each skill feeds into the next, so every step knows what came before.
The catches
- Maturity: the README says gstack runs on more agents than Claude, but the experimental tier has no certification run yet, and on those hosts the safety skills are advisory, not blocked.
- Platforms: Windows and Gemini models get minimal support, and Windows 11 with Smart App Control on refuses to start the compiled binaries.
- Privacy: telemetry is off by default, but outside reviews are on by default and send code to Codex or Claude Code under your own login.
- Depth: without Docker, security audits run static assessment only.
- Evidence: the productivity story sits in a separate document, which we did not read.
What we did not test
We did not install gstack, run any of its commands or check the review, QA or security skills. We did not read the linked documents either. Every claim here comes from a literal quote in the project's README or the GitHub API, and the feature descriptions are the project's own.
Verdict
Editors' view: this suits people who already use Claude Code and want a ready-made process around it, with planning, review and release steps. If you work on Windows or with another agent, expect rougher edges. Try it on a small project first, and check what the outside reviews send.
Sources
Need help adopting this, or have a software project in mind?
Tell us what you are building. The initial consultation is free. Email [email protected] or follow @zolnaltd.
Start a conversation →