aidcrew
Open-source CLI that runs a team of AI coding agents in parallel, each in its own isolated git worktree
🔗 Visit aidcrew
Description
Running one AI coding assistant at a time is fine for small edits, but a real feature often needs several tasks worked in parallel — refactor here, write tests there, review a PR somewhere else — without the agents stepping on each other's changes. aidcrew is a command-line tool that spins up several AI coding agents at once, each in its own isolated git worktree, so their work never collides.
Under the hood it's a TypeScript/Bun CLI with a plugin architecture: each agent job can target a different provider or model (Anthropic, Gemini, or OpenAI-compatible APIs like DeepSeek, Zen, OpenRouter), which lets you assign expensive models to planning and decisions while cheap models handle exploration or review. It ships a TUI to watch every agent's session side by side, built-in file/bash tools and inter-agent messaging, automatic session persistence, and cross-platform binaries for macOS, Linux, and Windows. It's free and open source (MIT-style), with 12 commits and active maintenance at time of review — you only pay for whatever provider API keys you plug in.
💬 Our review
The short version: aidcrew is a genuinely useful idea — isolate parallel coding agents in separate git worktrees so they can't corrupt each other's work — executed as a lean CLI rather than a heavyweight platform.
Compared to running Claude Code or Cursor's agent mode one session at a time, or heavier multi-agent frameworks like CrewAI/AutoGen that are built for scripted pipelines rather than interactive coding, aidcrew's niche is being provider-agnostic and git-worktree-native: you're not locked into one vendor, and isolation is structural (separate worktrees) rather than just separate chat tabs. It's early-stage — a handful of commits, small GitHub following — so expect rough edges and a project that could stall without a sustaining maintainer. Worth it if you already juggle multiple coding-agent sessions manually and want the isolation and provider-mixing handled for you; skip it if you only ever run one agent session at a time.
💰 Pricing
📊 Global score
🤖 AI-enriched data
MIT-style license, gratuit. Coûts réels = clés API des providers connectés (Anthropic, Gemini, OpenAI-compatibles).
Pros
Isolation structurelle par git worktree, pas juste des onglets séparés
Agnostique au provider (Anthropic, Gemini, OpenAI-compatibles)
TUI pour surveiller plusieurs agents en simultané
Gratuit et open source, binaires cross-platform
Cons
Projet jeune (12 commits), petite communauté
Pas de garantie de maintenance long terme
Demande de gérer soi-même plusieurs clés API