VBW (Vibe Better With Claude Code) is a Claude Code plugin, not a standalone tool. You install it withDocumentation Index
Fetch the complete documentation index at: https://mintlify.com/swt-labs/vibe-better-with-claude-code-vbw/llms.txt
Use this file to discover all available pages before exploring further.
/plugin install inside Claude Code, and it gives you a complete development lifecycle: phased planning, parallel agent execution, continuous verification, and persistent state across sessions.
VBW requires Claude Code v1.0.33+ and Opus 4.6+. It runs on macOS and Linux only. If you’re on Windows, use Claude Code inside WSL.
The core workflow
Every VBW project follows the same loop. You only need to remember two commands..vbw-planning/, configures Agent Teams, installs git hooks, and suggests relevant skills. For existing codebases, it auto-chains to /vbw:map first — running 4 Scout agents in parallel to analyze your code before planning begins.
.vbw-planning/ across sessions, so you can close your terminal and pick up exactly where you left off.
.vbw-planning/milestones/, and tags the git release.
That’s the entire lifecycle: /vbw:init → /vbw:vibe (repeat) → /vbw:vibe --archive.
Key capabilities
Agent teams
Seven specialized agents — Scout, Architect, Lead, Dev, QA, Debugger, and Docs — each with platform-enforced tool permissions that can’t be bypassed during compaction.
Native hooks
22 hooks across 11 event types run automatically for security filtering, commit validation, task gating, session lifecycle, and post-compaction verification.
Three-tier verification
Goal-backward QA runs after every phase. The
file-guard.sh hook blocks writes to files outside a Dev agent’s declared ownership. Database safety intercepts 40+ destructive patterns before they reach the shell.Real-time statusline
Five or six lines of situational awareness rendered after every response — phase progress, plan completion, effort profile, QA status, and agent health.
Session resumption
/vbw:resume reads ground truth directly from .vbw-planning/ files without requiring a prior /vbw:pause. It detects interrupted builds and tells you exactly what to do next.Skills.sh integration
Auto-detects your tech stack during
/vbw:init and recommends relevant skills from the open registry. Run /vbw:skills anytime to browse and install more.Token efficiency
Every new VBW capability runs as a bash subprocess at zero model token cost. Here’s how VBW’s coordination overhead compares to running stock Opus 4.6 agent teams manually:| Category | Stock agent teams | VBW | Saving |
|---|---|---|---|
| Base context overhead | 10,800 tokens | 1,500 tokens | 86% |
| State computation per command | 1,300 tokens | 200 tokens | 85% |
| Agent coordination (×4 agents) | 16,000 tokens | 1,200 tokens | 93% |
| Compaction recovery | 5,000 tokens | 700 tokens | 86% |
| Context duplication (shared files) | 16,500 tokens | 900 tokens | 95% |
| Total coordination overhead | 87,100 tokens | 12,100 tokens | 86% |
Platform requirements
- Operating system: macOS or Linux. Windows requires WSL.
- Claude Code: v1.0.33 or later (recommended: v2.1.47+ for full feature compatibility).
- Model: Opus 4.6 or later.
- Shell: bash (all hooks and scripts run as bash subprocesses).
jq: the only external dependency —brew install jq(macOS) orapt install jq(Linux).
Next step
Install VBW
Two commands inside Claude Code. No manual setup required.