supervibe

Supervibe

Supervibe is a local workflow layer for AI coding tools. It helps Claude Code, Codex CLI, Gemini CLI, Cursor, OpenCode, and GitHub Copilot CLI work from project context instead of chat-only guesses.

It runs locally. No Docker is required for normal setup.

v4.0 - current plugin v4.0.0. License: MIT.

Example startup signal: [supervibe] welcome plugin v4.0.0 initialized for this project.

Supervibe ships 114 skill bundles, project memory, CodeGraph, command routing, review gates, and release checks.

Compliance notice: Supervibe is for development assistance. You are responsible for following the Terms of Service and Acceptable Use Policy of every AI/provider service you use, including Anthropic when you use Claude.

Install

Requirements:

• Node.js 22.5 or newer, with node:sqlite
• Git
• Internet access during install so npm dependencies can be installed
• One supported local AI coding tool

macOS or Linux:

curl -fsSL https://raw.githubusercontent.com/vTRKA/supervibe/main/install.sh | bash

Windows PowerShell:

irm https://raw.githubusercontent.com/vTRKA/supervibe/main/install.ps1 | iex

Codex installs use the provider-scoped plugin root ~/.codex/plugins/marketplaces/supervibe-marketplace.

Then fully restart the AI tool.

If setup looks degraded after restart, run this in a terminal:

supervibe doctor --host all

Install and update integrity details live in Install integrity.

Release notes live in CHANGELOG. Current release evidence links Visible context intelligence, Performance SLOs, rollback guidance, and 561 test files.

First-Day Checklist

Use the skills-only path first. It keeps the startup route short while selected skill bundles, memory, CodeGraph, and a clear proof trail stay active.

1. Install Supervibe once.
2. Restart the AI tool.
3. Open a real project.
4. Type /supervibe-genesis in the AI chat and review the dry-run.
5. Approve only when the detected host, stack, managed files, memory, indexes, rules, and skill bundles look right.
6. Type /supervibe-status.
7. Type /supervibe for the next safe action.

After genesis, /supervibe is the normal entrypoint. If an active graph exists, it resumes from that graph and shows one visible next action. If there is no active graph, it routes to brainstorm, plan, design, audit, or maintenance. The first-day route is genesis -> status -> /supervibe, and smooth graph waves stay behind the visible command.

Where To Type Things

What you see	Where to type it	Example
Starts with /supervibe	AI chat	/supervibe-genesis
Starts with supervibe and no slash	Terminal	supervibe doctor --host codex
Starts with npm run	Terminal in the plugin checkout	npm run supervibe:status
Starts with node <resolved-supervibe-plugin-root>/...	Terminal	node <resolved-supervibe-plugin-root>/scripts/supervibe-status.mjs

Do not paste slash commands into PowerShell, bash, or zsh. Normal terminals do not understand /supervibe-* commands.

Main Workflow

For most feature work, start with /supervibe. When you need explicit manual commands, the public path is:

/supervibe-brainstorm "idea"
/supervibe-plan --loop-ready --from-brainstorm <spec-path>
/supervibe-plan --review <plan-path>
/supervibe-loop --atomize-plan <plan-path> --plan-review-passed
/supervibe-loop --guided --file <graph.json>
/supervibe-loop --validate-completion --file <graph.json>
/supervibe-review
/supervibe-ship

If the work needs UI direction, a mockup, or a prototype handoff, run /supervibe-design before planning and feed the approved output into the plan.

flowchart TD
  accTitle: Supervibe idea-to-release workflow
  accDescr: Brainstorm approves scope, optional design and prototype work feeds planning, reviewed plans become graph work, scoped verification happens before review, and ship completes release readiness.
  subgraph discover["Discover and approve"]
    idea["Idea"] --> brainstorm["Brainstorm"]
  end
  subgraph designLane["Optional design and prototype"]
    brainstorm -. "optional UI or prototype work" .-> design["Design direction"]
    design --> prototype["Prototype handoff"]
  end
  subgraph planGraph["Plan and graph"]
    brainstorm --> plan["Plan"]
    prototype --> plan
    plan --> planReview["Plan review"]
    planReview --> loopGraph["Loop graph"]
  end
  subgraph executeVerify["Execute and verify"]
    loopGraph --> graphWork["Graph work"]
    graphWork --> verify["Verify scoped evidence"]
  end
  subgraph release["Review and ship"]
    verify --> review["Review"]
    review --> ship["Ship release"]
  end

Plain contract:

idea -> brainstorm -> optional design/prototype -> plan -> plan review -> loop graph -> graph work -> verify scoped evidence -> review -> ship release