How can AI coding support become more reliable? There’s a way to separate product planning, engineering reviews, releases, and QA into clear operational modes. This is what Garry Tan has released as an open-source toolkitgstackpackages Claude Code into 8 professional workflow techniques, providing a persistent browser runtime. gstack isn’t just about adding a new model layer; it’s designed to help Claude Code operate more efficiently by clearly separating roles in product planning, engineering reviews, releases, and testing processes.

The emergence of AI coding tools has recently increased expectations for increased developer productivity. However, it’s important to maintain product quality and efficiently manage the deployment process, not just generate code. At this pointworkflowmanagement becomes important, and gstack can be considered an innovative attempt to solve this.

8 core commands

The gstack repository currently provides 8 major commands./plan-ceo-review(CEO review plan),/plan-eng-review(engineering review plan),/review(code review),/ship(release),/browse(browser browsing),/qa(QA testing),/setup-browser-cookies(browser cookie setup),/retro(retrospective) are available. Each command is linked to a specific operational mode and is designed to efficiently separate and manage software delivery tasks. Especiallyworkflowsystematically defines tasks to maximize work efficiency.

Persistent browser is a core system

The most important technical aspect of gstack is not the Markdown technology, but the browser subsystem. gstack provides Claude Code with a persistent browser environment, and maintaining this environment is the most challenging part. Instead of running a new browser for each task, gstack runs a long-running headless Chromium daemon and communicates via localhost HTTP. This helps reduce latency and maintain state. Cold starts take about 3-5 seconds, but subsequent calls after startup run within about 100-200ms. Because the browser is alive, cookies, tabs, localStorage, and login status are maintained between commands. Also, the server automatically shuts down if it is idle for 30 minutes. This persistent browser environment plays an important role in improving the efficiency ofworkflowprocesses.

How to connect browser automation to QA

This daemon architecture is very important for QA and browser-based development. In many agent workflows, browser automation is used as a debugging step or screenshot utility. In gstack, browser access is part of the core workflow./browsemode allows agents to log in, browse the app, take screenshots, and inspect errors./qaextends these features by analyzing browser diffs, identifying affected paths, and testing relevant pages or flows. This sample flow shows/qachecking 8 changed files and 3 affected paths, and then testing those paths for a local app instance. That is, the project attempts to connect source changes to actual application behavior, and does not treat QA as a separate manual step.workflowstrengthens this connection.

Installation requirements and project layout

The repository’s implementation choices are very specific. gstack requires Claude Code, Git, and Bun v1.0+. The package.json file currently shows version 0.3.3, lists Playwright and diff as runtime dependencies, and compiles the browse executable from the browse source tree. According to the repository README,/browsecompiles a native binary and supports macOS and Linux, x64 and arm64. The installation flow copies the repository to~/.claude/skills/gstackand runs./setupand registers the skill for Claude Code. The team can also copy the same setup to the project’s.claude/skills/gstackdirectory to shareworkflowprocesses.

Why use Bun

The architecture document explains why Bun is used instead of traditional Node.js setups like Node.js. Four reasons are specified: compiled binary, native SQLite access, native TypeScript execution, and a built-in HTTP server, Bun.serve(). These choices are not cosmetic but practical. gstack directly reads the Chromium SQLite cookie database, and Bun’s built-in database support eliminates the need for additional native packages. The compiled binary model also aligns with the repository’s installation style, and users are not expected to manage a separate runtime toolchain within~/.claude/skills/Therefore, these technologies were chosen to buildworkflowprocesses smoothly.

Key takeaways

• gstack is a workflow layer, not a new model or agent framework.
• A persistent browser daemon is a core technical component.
• QA is directly connected to code changes.
• gstack uses Bun for compiled binaries, native SQLite access, native TypeScript execution, and a built-in HTTP server for the browser daemon.
• gstack’s contribution is its operational structure. It narrows the scope of responsibility by separating product reviews, engineering reviews, code reviews, releases, and browser-based verification into clear modes.

In conclusion, gstack is an important tool for innovating Claude Code-powered developmentworkflowprocesses. It is expected to contribute to increased developer productivity and code quality in the future.