Welcome to The Weekly Five - your curated list of 5
exceptional open source projects I discovered this week.
cmux
If you're working with AI coding agents like Claude Code, Codex, or Gemini, you know the pain of juggling multiple terminal sessions and missing that crucial moment when your agent needs input. cmux is a Ghostty-based macOS terminal built specifically for this workflow , it keeps you notified when your AI agents need attention, so you can actually step away without anxiety.
Key Features
Notification rings: Panes get a blue ring and tabs light up when coding agents need your attention , no more missed prompts
Notification panel: See all pending notifications in one place and jump to the most recent unread with a single click
Vertical + horizontal tabs: Sidebar displays git branch, linked PR status/number, working directory, listening ports, and latest notification text
In-app browser: Split a browser alongside your terminal with a scriptable API ported from Vercel's agent-browser
Multi-language support: README available in 20+ languages including Japanese, Chinese, Korean, Spanish, and more
How to Use
Download the latest release from the macOS DMG directly from the releases page
Drag cmux to your Applications folder and launch it
Start your AI coding agent (Claude Code, Codex, OpenCode, etc.) in a new pane , cmux will automatically detect when it needs your attention
Use the vertical tabs sidebar to manage multiple agent sessions and monitor their status at a glance
cmux has quickly become a favorite in the AI-assisted development community, with 13,000 stars, 78 contributors, and 919 forks on GitHub. There are currently 1,099 open issues, which speaks to an active community pushing for new features and improvements.
superpowers
Superpowers is a complete software development workflow framework that transforms how your coding agents approach building software. Instead of letting AI jump straight into writing code, it enforces a structured methodology,starting with proper requirements gathering, breaking specs into digestible chunks, and then executing via test-driven development with subagent coordination.
Key Features
Spec-first approach: Your agent asks clarifying questions and builds a proper specification before touching any code
Subagent-driven development: Launches multiple agents to work through engineering tasks while inspecting and reviewing their work,enabling hours of autonomous operation
Built-in best practices: Emphasizes true red/green TDD, YAGNI (You Aren't Gonna Need It), and DRY principles throughout the workflow
Automatic skill triggering: Skills activate contextually without requiring special commands,your agent just works smarter
Multi-platform support: Works with Claude Code, Cursor, Codex, and OpenCode through their respective plugin systems
How to Use
For Claude Code users: Run
/plugin marketplace add obra/superpowers-marketplacethen/plugin install superpowers@superpowers-marketplace(or use the official marketplace with/plugin install superpowers@claude-plugins-official)For Cursor users: Simply run
/add-plugin superpowersin the Agent chat or search the plugin marketplaceFor Codex/OpenCode users: Tell your agent to fetch and follow instructions from the appropriate install URL in the repo's docs folder
Start coding: Once installed, skills trigger automatically,just begin your project and the structured workflow kicks in
With over 138,000 stars, 11,749 forks, and 31 contributors actively developing the framework, Superpowers has clearly struck a chord with developers looking to make their AI coding assistants more reliable and methodical.
everything-claude-code
If you've been using AI coding agents like Claude Code, Codex, or Cursor, you know they're powerful out of the box , but they can be so much more with the right setup. Everything Claude Code is a comprehensive performance optimization system born from an Anthropic hackathon winner, offering production-ready configurations, skills, memory optimization, and security scanning refined over 10+ months of daily real-world use.
Key Features
Universal agent support , works across Claude Code, Codex, Cursor, OpenCode, Gemini, and other AI agent harnesses
Complete optimization system , not just configs, but skills, instincts, memory optimization, and continuous learning baked in
Security-first approach , includes built-in security scanning via the ecc-agentshield package
Multi-language ecosystem , supports 12+ language ecosystems including TypeScript, Python, Go, Java, and more
Research-first development , configurations and patterns evolved from building real products
How to Use
Install the universal package via npm:
npm install ecc-universalFor security scanning capabilities, add:
npm install ecc-agentshieldConfigure your preferred AI agent harness (Claude Code, Cursor, etc.) using the provided MCP configurations and rules
Check out the companion guides on X/Twitter for detailed walkthroughs on getting the most from the system
With over 144K stars, 22K+ forks, and 159 contributors, this is one of the most actively developed AI agent optimization projects in the open source community , and with only 72 open issues, it's clear the maintainers are keeping things tight.
t3code
T3 Code is a minimal web GUI that lets you interact with coding agents like Codex and Claude through a clean, unified interface. If you've been juggling multiple AI coding tools and want a simpler way to work with them, this project from the Ping team (yes, the create-t3-app folks) is worth checking out.
Key Features
Supports multiple coding agents (Codex and Claude) with more coming soon
Run instantly without installation via
npx t3Native desktop apps available for Windows, macOS, and Arch Linux
Lightweight, minimal interface focused on productivity
Easy authentication flow that piggybacks on existing CLI tools
How to Use
First, install and authenticate at least one supported provider , either Codex CLI (run
codex login) or Claude Code (runclaude auth login).Run
npx t3to launch the web GUI instantly without any installation.Alternatively, install the desktop app via your preferred package manager:
winget install T3Tools.T3Codeon Windows,brew install --cask t3-codeon macOS, oryay -S t3code-binon Arch Linux.
The project is still early days , the maintainers are upfront about expecting bugs and aren't accepting contributions just yet. But with 8,450 stars, 90 contributors, and an active Discord community, T3 Code has clearly struck a chord with developers looking for a better coding agent experience.
zeroclaw
ZeroClaw is a lightning-fast, fully autonomous AI personal assistant infrastructure written entirely in Rust. What makes it stand out? It can run on hardware as cheap as $10 with less than 5MB of RAM , that's 99% less memory than alternatives like OpenClaw. Built by students and members from Harvard, MIT, and Sundai.Club, it's designed to deploy anywhere and swap out any component you need.
Key Features
Ultra-lightweight: Runs on minimal hardware with <5MB RAM, making it accessible for edge devices and budget setups
100% Rust: Zero overhead, zero compromise , built for performance and safety from the ground up
Platform agnostic: Works on any OS, any platform , deploy wherever you need it
Fully autonomous: Designed as agentic AI infrastructure that operates independently
Modular architecture: Swap out components freely to customize your assistant stack
How to Use
Install Rust (if you haven't already): Visit rust-lang.org to get the latest Rust toolchain (Edition 2024 recommended).
Clone the repository:
git clone https://github.com/zeroclaw-labs/zeroclaw.gitBuild the project: Navigate to the directory and run
cargo build --releaseRun ZeroClaw: Launch your personal AI assistant with the compiled binary and configure it for your specific use case.
With nearly 30K stars, 239 contributors, and 229 open issues showing active development, ZeroClaw has built a thriving community around making AI assistants accessible to everyone , not just those with expensive hardware.
If you are doing Open Source I have a good news for you, I work at CodeRabbit which is an AI review tool and its free for Open Source, please reach out to me on X or LinkedIn or just send an email on [email protected] if you need help on adopting CodeRabbit.
You can visit our portal below to create a new account and connect your repository and start reviewing your code.
