📈 swarm-forge · ⭐ 2,902 (+1,579 this month)
SwarmForge is a disciplined tmux-based agent orchestration platform that coordinates multiple AI agents working on the same project. It provides shared operational scripts, role-specific prompts, and message passing to enable agents to collaborate without conflicts. The system supports configurable workflows via branches, such as two-pack, four-pack, and six-pack, each with distinct agent roles and quality gates.
• Worktree-based agent collaboration
• Role-specific prompt configurations
• Multiple workflow presets included
#AI #Agents #DevTools
🔥 andrej-karpathy-skills · ⭐ 206,188 (+491 today)
multica-ai/andrej-karpathy-skills provides a CLAUDE.md file with four principles—Think Before Coding, Simplicity First, Surgical Changes, and Goal-Driven Execution—to guide Claude Code toward more reliable behavior. It directly addresses common LLM coding issues like wrong assumptions, overengineering, and unnecessary edits. The file is designed to be dropped into a project to improve agent output quality.
• Encodes four actionable coding principles
• Targets common LLM coding pitfalls
• Single-file drop-in configuration
#AI #ClaudeCode #LLM
5/ plugins — Ships official Cursor plugins for developer tools, frameworks, and SaaS products 👉 https://t.co/axUibcMXc3 ⭐ 4,891 total · +1,761 this week
🏆 Top 5 trending AI repos on GitHub this week 🧵👇
1. MoneyPrinterTurbo — one-click AI short video generator
2. OpenViking — self-evolving context database for AI agents
3. ai-memory — long-term memory for coding agents
4. modular — unified AI dev platform
5. plugins — official Cursor plugin collection
4/ modular — Combines the Mojo language and MAX framework for building and deploying AI models 👉 https://t.co/BAIaW62oSg ⭐ 29,035 total · +2,176 this week
📈 deepseek-harness · ⭐ 178,465 (+92,686 this week)
DeepSeek Harness (dsh) is an open-source agent harness from DeepSeek AI that adopts an everything-is-a-plugin architecture, powered by the Cordis framework. It is currently in developer preview, with rapid iteration and potential breaking changes. The harness runs via npm or source, providing a web UI for local or SSH launches.
• Built on Cordis framework
• Runs via npm or source
• Web UI for local or SSH
#AI #LLM #Agents
🔥 sub2api · ⭐ 38,693 (+673 today)
Sub2API is an open-source API gateway that consolidates multiple AI subscription services—Claude, OpenAI, Gemini, and Grok—into a single unified interface. It enables quota sharing among users to reduce costs and integrates natively with tools like Claude Code and Codex. The project includes a clear disclaimer about potential terms-of-service risks and is intended for technical learning and research.
• Supports Claude, OpenAI, Gemini, Grok
• Enables shared subscription quota pooling
• Works with Claude Code and Codex
#AI #API #LLM
🔥 codex · ⭐ 112,209 (+4,159 today)
OpenAI Codex CLI is a lightweight coding agent that runs directly in your terminal, bringing AI-assisted development to the command line. It supports sign-in with ChatGPT plans or API keys, and can be installed via curl, npm, or Homebrew on Mac, Linux, and Windows. The tool is open-sourced under Apache-2.0 and integrates with IDEs like VS Code, Cursor, and Windsurf.
• Install via curl, npm, or Homebrew
• Works with ChatGPT or API key
• Open-source under Apache-2.0
#AI #CodingAgent #OpenSource
🔥 substrate · ⭐ 1,537 (+245 today)
Agent Substrate is a control plane for deploying and managing large-scale agent workloads. It maps many idle agents onto fewer physical workers via suspend/resume and full-state snapshots, achieving heavy multiplexing. It supports microVMs and gVisor, integrates with Kubernetes, and is framework-agnostic, hosting agents built on ADK, LangChain, or other stacks.
• sub-second agent suspend/resume
• 30x+ oversubscription via multiplexing
• framework-agnostic OCI container support
#AI #Agents #Infrastructure
🔥 AI-Infra-Guard · ⭐ 4,792 (+28 today)
AI-Infra-Guard is a full-stack AI red teaming platform from Tencent Zhuque Lab. It integrates Agent Scan, Skills Scan, MCP scan, AI Infra vulnerability scan, and LLM jailbreak evaluation to help users self-examine AI security risks. The platform includes a frontend, standalone CLIs, a skill market, and a large vulnerability rule library.
• Agent, MCP, and skill scanning
• Multi-turn jailbreak attack evaluation
• 2000+ CVE vulnerability rules
#AISecurity #RedTeaming #LLM