Vibe Coding Discover：AI 开源项目目录，收录 Agents、MCP、Skills、RAG

Discover top AI open source — Agents, MCP, Skills, RAG, Tools and Frameworks. Vibe Coding Discover helps you find the right project so you can code faster.

Today (10/05/2026 UTC): +14 projects, +79 use cases

Today's Recommendations

1

ponytail

★ 151K

Ponytail is a cross-agent skill/plugin that makes coding agents apply a YAGNI ladder (skip, reuse, stdlib, native, one-liner) before writing code. Its benchmarks claim ~54% less code and ~20% lower cost on Claude Code sessions, while keeping validation, security, and accessibility intact. Ships Claude Code, Codex, Copi

Skills | JavaScript · agent-skills · claude-code

View Project →

2

impeccable

★ 75K

Impeccable is a design skill pack for AI coding agents: 1 skill, 24 commands (/impeccable audit, polish, critique, animate), plus 61 deterministic detector rules and a browser extension that flag generic AI frontend design. Installs into Claude Code, Cursor, Codex, Copilot and more; the CLI rules run without an LLM.

Skills | JavaScript · design · ai-coding-agents

View Project →

3

Codewhale

★ 41K

A Rust terminal coding agent that reads repositories, edits files, runs commands, and checks its work using hosted or local models. Supports agent teams, MCP servers, skills, computer-use tools, and interactive or scripted workflows.

AI Coding | Rust · coding-agent · terminal

View Project →

tenuo

★ 101

Tenuo provides cryptographically signed, task-scoped warrants that constrain AI agent tools and arguments. Authorization is checked at execution boundaries, and delegated warrants can only narrow authority.

AI Agents | Rust · agent-security · authorization

View Project →

PentestGPT

★ 16K

Python agent for authorized penetration testing and CTFs. Its staged pipeline handles reconnaissance, exploitation, and reporting, using Claude Code or Codex; a separate interactive mode supports multiple LLM providers.

AI Agents | Python · penetration testing · security

View Project →

langchaingo

★ 9.7K

A Go implementation of LangChain for composing LLM applications. It provides integrations for model providers, agents, chains, embeddings, and vector stores, including components for retrieval-augmented question answering.

AI Frameworks | Go · LLM applications · agents

View Project →

skills

★ 978

A reusable skill pack for AI coding agents that supports feature discovery, spec writing, implementation, research, and code and visual review. Install skills into compatible harnesses such as Claude Code, Codex, opencode, and Cursor.

Skills | TypeScript · agent-skills · software-engineering

View Project →

syncso

★ 217

Provides agents with live New York events and activities through an MCP server and installable skill. Its tools find things to do and return details such as dates, prices, images, and booking links.

MCP | MCP server · agent skill

View Project →

claude-inspector

★ 131

An Electron desktop app that proxies Claude Code CLI traffic and displays API payloads, token costs, and session flows. It can use the local Claude Code CLI to analyze captured sessions and generate diagrams.

AI Tools | HTML · Claude Code · prompt inspection

View Project →

Codewhale

★ 41K

A Rust terminal coding agent that reads repositories, edits files, runs commands, and checks its work using hosted or local models. Supports agent teams, MCP servers, skills, computer-use tools, and interactive or scripted workflows.

AI Coding | Rust · coding-agent · terminal

View Project →

LangAlpha

★ 248

An early-stage multi-agent workflow for equity research. LangGraph agents gather financial data and news, perform analysis, and produce Markdown reports; it uses sources such as Polygon, Yahoo Finance, and Tavily.

AI Agents | Jupyter Notebook · multi-agent · financial research

View Project →

Daily Claim

Have an AI Open Source Project?

Submit a GitHub repository or an X recommendation. No login required.

Submit Project