dev
5 min read
GitHub Trending Daily Report — 2026-03-01
Today's GitHub Trending was dominated overwhelmingly by the Claude·AI agent ecosystem. Coding agents, multi-agent orchestration, RAG/LLM apps, and edge AI — practical AI tooling is the core trend of the day.
🔥 Today at a Glance — Top 3 Themes
| # | Theme | Representative Repos | Highlights |
|---|---|---|---|
| 1 | Claude Agent Ecosystem | claude-code, ruflo, superpowers | Claude-based coding & orchestration tools surging |
| 2 | AI Agent Frameworks | deer-flow, hermes-agent, OpenSandbox | Multi-agent & sandbox infrastructure spreading fast |
| 3 | LLM Apps & RAG Collections | awesome-llm-apps, MobileAgent | Explosive demand for practical LLM app references |
📅 Weekly Context — Changes vs. Last 7 Days
- Claude Ecosystem Dominance: anthropic/claude-code, ruvnet/ruflo, obra/superpowers and other Claude-centric agent dev tools keep reappearing. The Anthropic-led ecosystem is growing faster than OpenAI's.
- Surge in Agent Orchestration Demand: Multi-agent workflows and automation platforms are becoming mainstream, moving well beyond simple chatbots.
- MCP (Model Context Protocol) Expansion: Projects like datagouv/datagouv-mcp show MCP servers being built across diverse domains — now connecting open data and government data.
- Edge AI / Speech Recognition: Growing interest in lightweight ASR models for edge devices, like moonshine-ai/moonshine.
- Language Distribution: Python > TypeScript > Shell > C > Go. Python dominates AI/ML; TypeScript leads in agent UI and tooling.
🌟 Top 10 Repositories to Watch
1. obra/superpowers — ⭐ 1,314 stars today
- Language: Shell | Total stars: 65,754
- One-liner: An agent skill framework & software development methodology that works alongside Claude Code
- Why it's trending: The exploding popularity of Claude Code is driving demand for concrete methodologies to apply it in real work
- Application idea: Define repetitive tasks (writing tests, generating PR descriptions) as superpowers skills to standardize across your team
2. moeru-ai/airi — ⭐ 1,088 stars today
- Language: TypeScript | Total stars: 19,193
- One-liner: Self-hosted AI companion (waifu character) with real-time voice chat, Minecraft & Factorio play support
- Why it's trending: Community fascination with self-hosting Neuro-sama-style AI characters is exploding
- Application idea: Useful as a reference for analyzing local LLM + real-time TTS pipeline architecture
3. ruvnet/ruflo — ⭐ 938 stars today
- Language: TypeScript | Total stars: 16,352
- One-liner: Multi-agent swarm orchestration platform for Claude — with RAG & distributed intelligence support
- Why it's trending: Native Claude Code integration + enterprise architecture makes it easy to adopt in production
- Application idea: Compose multiple specialized agents (search, coding, review) into a swarm to automate complex tasks
4. bytedance/deer-flow — ⭐ 891 stars today
- Language: Python | Total stars: 22,526
- One-liner: Open-source SuperAgent for research, coding, and content generation — with sandbox, memory, and sub-agent support
- Why it's trending: ByteDance's open release lends credibility, and its ability to handle long-running tasks (minutes~hours) is drawing attention
- Application idea: Automate a data research → code writing → report generation pipeline
5. anthropics/claude-code — ⭐ 705 stars today
- Language: Shell | Total stars: 71,721
- One-liner: Agentic coding tool running in the terminal — understand, modify codebases and handle git workflows via natural language
- Why it's trending: Rapidly becoming the developer standard as Anthropic's official tool
- Application idea: Immediately applicable to understanding legacy codebases, iterative refactoring, and automating PR descriptions
6. Shubhamsaboo/awesome-llm-apps — ⭐ 605 stars today
- Language: Python | Total stars: 98,137
- One-liner: A collection of LLM apps, AI agents, and RAG examples built with OpenAI, Anthropic, Gemini, and open-source models
- Why it's trending: Ongoing demand for practical examples that enable rapid prototyping
- Application idea: Find a template for your desired use case and kick off a PoC immediately
7. moonshine-ai/moonshine — ⭐ 494 stars today
- Language: C | Total stars: 6,158
- One-liner: Fast and accurate automatic speech recognition (ASR) for edge devices
- Why it's trending: Growing demand for on-device AI; privacy-friendly ASR with no cloud dependency
- Application idea: Apply to a local meeting transcription & subtitle generation pipeline
8. alibaba/OpenSandbox — ⭐ 350 stars today
- Language: Python | Total stars: 1,740
- One-liner: Universal sandbox platform for AI apps — multi-language SDK, Docker/K8s runtime, coding agent, GUI agent & RL training support
- Why it's trending: An open-source alternative to build your own agent evaluation and execution environment
- Application idea: Safely isolate agent code execution environments to eliminate production risks
9. NousResearch/hermes-agent — ⭐ 183 stars today
- Language: Python | Total stars: 1,055
- One-liner: An agent framework based on the Hermes model
- Why it's trending: Expanding the open-source LLM ecosystem from Nous Research
- Application idea: Reference architecture when building your own agent on open-source models
10. datagouv/datagouv-mcp — ⭐ 122 stars today
- Language: Python | Total stars: 615
- One-liner: Official MCP server connecting France's national open data platform to AI chatbots
- Why it's trending: First MCP adoption by a government agency — a landmark for connecting public data to AI
- Application idea: Benchmark this to prototype a local MCP server connecting Korea's public data portal (or internal Notion DB, Slack, etc.)
📊 Language & Domain Distribution
| Language | Repo Count | Primary Use |
|---|---|---|
| Python | 7 | AI agents, RAG, sandbox, ASR |
| TypeScript | 4 | Agent orchestration, AI UI |
| Shell | 2 | Agent skills, coding tools |
| C / C++ | 2 | Edge AI, ML frameworks |
| Go / JavaScript | 2 | API relay, service infrastructure |
Domain: AI Agents (10) > LLM Tooling (3) > Infrastructure/Sandbox (2) > MCP (1)
✅ Next Actions — 3 Things to Try This Week
anthropics/claude-code+obra/superpowerscombo hands-on — Experience agentic coding workflows in the terminal firsthand, and try automating repetitive tasks with the superpowers skill systemruvnet/ruflomulti-agent architecture deep dive — Understand the swarm intelligence structure and design which tasks in real projects can be delegated to agents- Build your own MCP server — Use
datagouv/datagouv-mcpas a reference to prototype a local MCP server connecting frequently-used internal data sources (Notion DB, Slack, etc.)