vs gstack

buildcrew vs gstack

buildcrewgstack
Installnpx buildcrew (interactive setup)git clone + Bun + ./setup
OrchestrationAuto (orchestrator routes to agents)Manual (/qa, /review, ...)
Harnessinit + add + custom .mdManual CLAUDE.md
DependenciesNoneBun, Playwright binary (~58MB)
Browser testingPlaywright MCP (required)Custom Playwright daemon
Pipeline docsAuto-generated chain (01-07)Per-skill output
Agents18 (9 opus, 9 sonnet) + 7 Agent Skills34 skills
Model strategyMixed (opus for thinking, sonnet for doing)Same model for all
Second opinionBuilt-in (Codex or Claude subagent)Built-in (Codex)
Product thinkingthinker agent (6 forcing questions)/office-hours skill
Architecture reviewarchitect agent (before code)/plan-eng-review skill
Design reviewdesign-reviewer agent (8-dim scoring)/plan-design-review skill
CustomizationEdit .md directly, version auto-updateMay overwrite on update
Multi-agent supportClaude Code (subagents) + Agent Skills tools (Codex, Cursor, Copilot, Gemini CLI)Claude, Codex, Gemini CLI, Cursor

buildcrew strengths

  • Auto-orchestration β€” describe what you need, orchestrator routes to the right agents
  • Full lifecycle β€” think, plan, design, build, test, review, ship, monitor
  • Harness engineering β€” project context baked into every agent
  • Pipeline documentation β€” every decision traceable
  • Second opinion β€” independent review after every mode
  • Zero deps β€” just Markdown files
  • Interactive setup β€” one command, everything configured

gstack strengths

  • More skills β€” 34 vs 7 (buildcrew's 7 skills wrap 13 modes)
  • Multi-host β€” works with Codex, Gemini CLI, Cursor
  • Chrome cookie import β€” login as real user for QA
  • Community β€” 60K+ stars, active development
  • Design generation β€” AI-generated visual mockups (GPT Image API)
  • Performance benchmarks β€” Core Web Vitals regression detection

When to use which

  • buildcrew β€” you want one tool that handles everything automatically. Just talk to @buildcrew.
  • gstack β€” you want granular control with explicit /commands, or you use multiple AI tools beyond Claude Code.
  • Both β€” gstack for planning/review decisions, buildcrew for automated execution. They complement each other.