Run 100+ finished AI agents and RAG apps for free
You can run finished code with the features you need in 30 seconds without designing complex AI agents from scratch. This repository is a collection of over 100 open-source projects, including agent skills, standalone apps, and RAG pipelines that integrate with various AI models. After cloning the code, you can immediately experience diverse AI tasks, from data analysis to web search and voice conversation, by running simple commands.
Giving new skills to coding agents in 10 seconds
The fastest entry point to this repository is skills for coding agents. You can install a skill with specific functionality using the npx skills add command. For example, installing the 'Project Graveyard' skill gives your coding agent the ability to find abandoned side projects and analyze why they failed. All skills include actual code and have passed security and evaluation CI gates. They are ready to use immediately in various coding agents such as Claude Code, Codex, and Cursor.
Step-by-step agent composition from beginner to advanced
The repository categorizes projects by user proficiency. 'Starter AI Agents' are single-file agents that run with just an API key, suitable for beginners for tasks like converting blogs to podcasts or analyzing CSV files in natural language. 'Advanced AI Agents' are production-level agents equipped with tool calling, memory, and multi-step reasoning, handling complex tasks such as financial analysis or system architecture reviews. Additionally, 'Always-on Agents' run in the background based on schedules or events, automatically performing repetitive tasks like Hacker News briefings or dependency release monitoring.
Scope of RAG and multi-agent teams
Retrieval-Augmented Generation (RAG) pipelines cover a wide range, from simple chains to agentic RAG and multi-source search. Options vary from Llama 3.2-based RAG that operates fully offline in a local environment to GPT-4o-based autonomous RAG that falls back to web search. The multi-agent team section shows examples of multiple agents collaborating to perform complex tasks that cross domain boundaries, such as legal document analysis, game design, and hiring processes. Each team is designed with a structure that is not dependent on specific frameworks or models and can be combined as needed.
Models and costs to check before running
Most agents are compatible with various open-source and commercial models, including Claude, Gemini, GPT, DeepSeek, Llama, and Qwen. However, certain agents rely on specific model features (e.g., Gemini's real-time voice API, DeepSeek's reasoning capabilities), so you must verify that the model you intend to use supports those features. Some projects run locally and incur no API costs, but cloud-based agents may generate API call fees. Although the Apache-2.0 license allows commercial use, it is recommended to check the hardware specifications and model access permissions required by each agent in advance.
