A CLI and workflow engine that runs coding agents like Claude Code, Codex and Copilot through configurable review, repair and test gates before work is accepted.
mtarcure/claude-vibe-squad
Multi-model AI coding orchestrator where a coordinator routes Markdown-defined specialist tasks to Claude, Codex, Gemini, Kimi and Grok CLIs in isolated git worktrees.
- Stars
- 165
- Forks
- 26
- Open issues
- 0
- Last commit
- 30 days ago
What is it?
Vibe Squad is a macOS-first orchestration layer for AI coding work. You describe a goal to a coordinator called Chrono, which splits it into scoped task files, assigns each to one of 71 role-based specialists, and launches the matching native CLI for Claude, Codex, Gemini (via agy), Kimi, or Grok. Each mutating task runs in its own git worktree with a declared write scope, and results are integrated only after checks pass.
The behaviour lives in Markdown: modes, specialist briefs, capability cards and skills are plain files that can be read and edited. Code handles launching, worktree isolation, identity checks and integration boundaries. A reviewer from a different model family can be required for higher-risk tasks, and run notes are stored in a private Markdown vault outside the repository.
The project needs tmux, fswatch, jq, Python 3.13 and authenticated CLIs for each provider. Its docs state that agents run with your user's permissions and that write scope is enforced at integration rather than as a sandbox.
Key features
- 01Routes tasks to specialists bound to Claude, Codex, Gemini, Kimi or Grok CLIs
- 02Runs each mutating task in an isolated git worktree with a declared write scope
- 03Requires cross-family review for security- and judgment-critical work
- 04Stores specialist briefs, modes and skills as editable Markdown files
- 05Keeps run memory in a private Markdown vault outside the repository
Use cases
- Running several AI coding agents in parallel on separate branches of one repository
- Adding a second model family as an independent reviewer before merging AI-written changes
- Customising agent roles and workflows by editing Markdown instead of code
Quick start
- 1
Create a private memory vault and export its path
mkdir -p "$HOME/Obsidian-Chrono" && printf '%s\n' '{"vault_id":"my-private-vault","schema_version":1}' > "$HOME/Obsidian-Chrono/.chrono-vault" && export CHRONO_VAULT_ROOT="$HOME/Obsidian-Chrono" - 2
Clone the repository and install the Python environment
git clone https://github.com/mtarcure/claude-vibe-squad.git && cd claude-vibe-squad && uv sync - 3
Install the local leak guard hook and run the health check
bash docs/install/install-pre-commit-hook.sh && bin/squad doctor - 4
Launch the tmux control room with Chrono
bin/squad up
Things to know
- macOS only for now.
- Requires several authenticated native CLIs for Claude, Codex, Gemini (agy), Kimi and Grok.
- Agents run code from your checkout with your permissions; write scopes are not a sandbox.
- Some optional features such as chrono-dedup are not distributed and need separate setup.
Topics
FAQ
- What is claude-vibe-squad?
- Multi-model AI coding orchestrator where a coordinator routes Markdown-defined specialist tasks to Claude, Codex, Gemini, Kimi and Grok CLIs in isolated git worktrees. Vibe Squad is a macOS-first orchestration layer for AI coding work. You describe a goal to a coordinator called Chrono, which splits it into scoped task files, assigns each to one of 71 role-based specialists, and launches the matching native CLI for Claude, Codex, Gemini (via agy), Kimi, or Grok. Each mutating task runs in its own git worktree with a declared write scope, and results are integrated only after checks pass.
- Is claude-vibe-squad free to use? What license does it have?
- claude-vibe-squad is open source under the MIT license. Check the license terms before using it in commercial products.
- Is claude-vibe-squad still maintained?
- Actively maintained. The last commit to claude-vibe-squad was on Sep 21, 2026.
- What can I use claude-vibe-squad for?
- Running several AI coding agents in parallel on separate branches of one repository Adding a second model family as an independent reviewer before merging AI-written changes Customising agent roles and workflows by editing Markdown instead of code
- How do I install claude-vibe-squad?
- Create a private memory vault and export its path: mkdir -p "$HOME/Obsidian-Chrono" && printf '%s\n' '{"vault_id":"my-private-vault","schema_version":1}' > "$HOME/Obsidian-Chrono/.chrono-vault" && export CHRONO_VAULT_ROOT="$HOME/Obsidian-Chrono" Clone the repository and install the Python environment: git clone https://github.com/mtarcure/claude-vibe-squad.git && cd claude-vibe-squad && uv sync Install the local leak guard hook and run the health check: bash docs/install/install-pre-commit-hook.sh && bin/squad doctor Launch the tmux control room with Chrono: bin/squad up
- Where can I get claude-vibe-squad?
- The source code and installation instructions are on GitHub: https://github.com/mtarcure/claude-vibe-squad
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
A TypeScript CLI and terminal UI for defining, launching and running persistent teams of Claude Code and Codex agents in tmux.
A Claude Code plugin with self-correcting memory, persistent FTS5 research wikis, and workflow skills, commands and hooks for AI coding agents.
A web workspace where teams and AI coding agents share live spatial boards, running each agent on its own isolated git branch.
A Rust terminal workspace that runs coding agents in persistent panes, tracks which ones are working, blocked, or idle, and lets agents control it via CLI and socket API.
Desktop, mobile, web and CLI app for running and orchestrating multiple coding agents such as Claude Code, Codex and OpenCode in parallel worktrees.