gstack: Garry Tan's Claude Code setup, 23 tools
gstack is an MIT licensed TypeScript toolkit of 23 Claude Code tools covering CEO, design, engineering management, releases, docs and QA.
TL;DR: gstack is an MIT licensed TypeScript repository with 135,174 stars and 20,092 forks that packages one person’s Claude Code setup into 23 tools, assigned to roles like CEO, Designer, Eng Manager, Release Manager, Doc Engineer and QA. It is the highest-starred repository in today’s trending batch, by a wide margin.
What the project is
The description is unambiguous about what kind of artifact this is: Use Garry Tan’s exact Claude Code setup. The claim is not that gstack is a general agent framework, but that it is a specific working configuration, reproduced and packaged.
| Field | Value |
|---|---|
| Full name | garrytan/gstack |
| Description | Use Garry Tan’s exact Claude Code setup: 23 opinionated tools that serve as CEO, Designer, Eng Manager, Release Manager, Doc Engineer, and QA |
| Stars | 135,174 |
| Forks | 20,092 |
| Primary language | TypeScript |
| License | MIT |
| Topics | none published |
| Homepage | none listed |
| Created | 2026-03-11 |
| Last push | 2026-10-05 |
| Owner type | User |
The number to sit with is 20,092 forks against 135,174 stars, roughly one fork per seven stars. That is a very high cloning rate, and it points at people who are reading the implementation closely enough to modify it, which is what you would expect from a setup you are meant to adopt and then make your own.
The repository has no published topics and no homepage, which makes the description and the source the entire public story. A wiki is enabled, so documentation exists outside the repository metadata.
Six roles and one assistant
The 23 tools are organised by role, and the roles are the interesting part. CEO, Designer, Eng Manager, Release Manager, Doc Engineer and QA is not a list of coding agents, it is a list of the people involved in shipping software, expressed as capabilities the assistant can be asked for.
That framing is a design choice with consequences. A tool that behaves as a release manager and a tool that behaves as a QA reviewer will want different inputs and enforce different things, and a single opinionated set cannot be right for every team’s process. The word opinionated in the description is doing the honest work there.
TypeScript as the primary language is worth noting for a Claude Code project, since it means the tools are installable through the Node ecosystem most developers already have.
How to look at it yourself
A clone is the fastest way to see what 23 tools actually looks like on disk, and given the fork rate, plenty of people start here:
git clone https://github.com/garrytan/gstack.git
cd gstack
ls
The description is unusually rich for a repository with no topics, so the API is worth hitting for the fields the description does not carry, like creation date and last push:
curl -s https://api.github.com/repos/garrytan/gstack \
| jq '{stars: .stargazers_count, forks: .forks_count, license: .license.spdx_id, created_at, pushed_at}'
When to use it
Treat gstack as a reference configuration rather than a dependency. The MIT license and the TypeScript implementation make it easy to read, copy the parts that match how your team works, and discard the rest. At 135,174 stars it is the clearest example in today’s trending batch of a personal agent setup being turned into a public artifact, and that is worth studying whether or not you adopt it.
Two things to check before installing. First, the role assignments are one person’s defaults, so the QA and release behaviours in particular are likely to encode assumptions about process that do not match yours. Second, a bundle of 23 tools is a lot of surface area to adopt blind. The wiki and the source are where to confirm what each tool does before you wire it into a workflow you depend on.
Source and Accuracy Notes
- Source repository: github.com/garrytan/gstack (verified via GitHub API
html_url) - License: MIT (verified via GitHub API
license.spdx_id) - Stars at time of writing: 135,174
- Forks at time of writing: 20,092
- Facts in this post come from the GitHub API repository endpoint and the repository description only. The README was not fetched, so the names of the 23 tools, install commands and their individual behaviours are not claimed here beyond the role list stated in the repository description.
Related Posts
ai-setup
Recall – Persistent Memory for Claude Code via MCP Hooks
Recall gives Claude Code a permanent memory store that survives session restarts and context compaction. Four hooks capture and restore context automatically — with cloud SaaS or self-hosted options.
2/28/2026
dev-tools
Automotive Skills Suite for AI Engineering
Evaluate Automotive Skills Suite for APQP, ASPICE, HARA, safety-plan, and DIA workflows with setup notes, governance risks, and SME review guidance.
5/28/2026
dev-tools
awesome-agentic-ai-zh Roadmap Guide
Explore awesome-agentic-ai-zh as a Chinese agentic AI learning roadmap, with setup notes, track selection, study workflow, and evaluation guidance.
5/28/2026