Awesome LLM Apps: 100+ Open-Source LLM Applications
Awesome LLM Apps is a hand-built collection of 100+ open-source AI agent, agent skill, and RAG app examples released under Apache-2.0. Reach for it when you want a working starting point to copy and adapt, not a library to import. Skip it if you want one polished framework instead of a grab-bag of standalone scripts with their own dependencies.
What is Awesome LLM Apps?
Awesome LLM Apps is a GitHub repository of 100+ open-source LLM applications: ready-to-run example code for AI agents, agent skills, and retrieval-augmented generation apps, released under the Apache-2.0 license. It spans single-file starter agents, multi-agent teams, voice agents, RAG pipelines, and crash-course tutorials for frameworks like Google ADK and the OpenAI Agents SDK.
Key Features and Application Categories
- ✓Agent Skills folder: install one with `npx skills add <github-url>` and use it in plain English; the README says each skill contains real code and successfully passes through a security + eval CI gate for Claude Code, Codex, and Cursor.
- ✓Starter AI Agents: single-file scripts that need only an API key, covering a travel planner, a blog-to-podcast converter, a medical imaging agent, and more.
- ✓Advanced AI Agents and Multi-agent Teams: production-style examples with tools, memory, and multi-step reasoning, including team examples like an AI Legal Agent Team and an AI Recruitment Agent Team.
- ✓RAG Tutorials: pipelines ranging from a Basic RAG Chain to Corrective RAG, Knowledge Graph RAG with Citations, and Multimodal Agentic RAG over text, PDFs, images, audio, and video.
- ✓Always-on and MCP Agents: scheduled background agents such as an always-on Hacker News briefing agent, plus agents wired to Model Context Protocol servers for GitHub, Notion, and browser control.
- ✓LLM Optimization and Fine-tuning: the Toonify example claims a 30-60% API cost cut via a TOON format, Headroom claims 50-90% via context optimization, and there are LoRA fine-tuning recipes for Gemma 3 and Llama 3.2.
- ✓Broad model coverage: according to the README, compatibility extends to examples for Claude, Gemini, GPT, DeepSeek, Llama, Qwen, and various other open-source models.
- ✓Framework Crash Courses: deep-dive tutorials such as the Google ADK Crash Course, covering tools, memory, callbacks, plugins, and multi-agent patterns.
Getting Started with an Agent or Skill
Two paths, per the README. For an Agent Skill: run `npx skills add https://github.com/Shubhamsaboo/awesome-llm-apps/tree/main/agent_skills/<name>`, which the README says takes about 10 seconds, then ask your coding agent to use it. For a full agent: `git clone https://github.com/Shubhamsaboo/awesome-llm-apps.git`, `cd` into a specific folder such as `starter_ai_agents/ai_travel_agent`, run `pip install -r requirements.txt`, then `streamlit run travel_agent.py` — the README times this at about 30 seconds. Each of the 100+ folders has its own requirements.txt and, in most cases, needs its own API key(s); there's no single global install.
Running and Customizing Examples
Usage is per-example: open the specific folder, read its own instructions, install its own requirements.txt, set whatever API key it calls for (OpenAI, Gemini, Anthropic, and so on, depending on the example), and run it — often via `streamlit run`, sometimes as a plain Python script. Customization means editing the source file directly, since these are standalone examples rather than a library with a documented import API. The top-level README lists the folders and one-line descriptions but doesn't detail per-example configuration beyond its one `ai_travel_agent` walkthrough, so budget time to read each example's own files before running it.
Practical Applications for LLM Agents
- •Prototype a niche agent fast: clone the AI travel agent or AI blog-to-podcast agent, then swap in your own prompts and tools.
- •Add a one-off automation to a coding agent via `npx skills add`, such as the Project Graveyard skill that reviews abandoned side projects.
- •Learn RAG patterns end to end, from the Basic RAG Chain through Corrective RAG to Knowledge Graph RAG with Citations, before designing your own pipeline.
- •Study multi-agent orchestration as a reference architecture, using examples like the AI Legal Agent Team or AI Recruitment Agent Team.
- •Run a fine-tuning recipe, such as the Gemma 3 or Llama 3.2 LoRA examples, as a template for tuning your own open-source model.
Advantages of Using This Collection
- ✓Every example is runnable, not pseudocode - the README says each is hand-built and tested end-to-end.
- ✓Apache-2.0 licensed, and the README states outright that you can clone, ship, and sell the code.
- ✓Agent Skills carry what the README calls a security + eval CI gate, a concrete check rather than just an unverified promise of quality.
- ✓Wide model coverage - Claude, Gemini, GPT, DeepSeek, Llama, and Qwen examples sit side by side, useful for comparing providers without a rewrite.
- ✓130,781 stars and 19,294 forks point to sustained community pickup, though that's a popularity signal, not a guarantee any one example is bug-free.
Considerations and Scope
- △No single global install: 100+ examples each carry their own requirements.txt and their own API-key needs, so setup cost is per-example, not one-time.
- △Not a library - there's no stable import surface or pip package; you copy source files and edit them directly.
- △Depth varies by folder: some are single-file scripts runnable in 30 seconds, others (multi-agent teams, MCP agents) assume more moving parts the top-level README doesn't spell out.
- △A few entries, like Openwork and the Wispr Flow-style voice dictation agent, link out to separate external repos rather than including the code here.
Alternative LLM Agent and RAG Resources
Frequently Asked Questions
Awesome LLM Apps is released under the Apache-2.0 license, and the README explicitly invites cloning, shipping, and even selling the code.
The Awesome LLM Apps README indicates its examples support models such as Claude, Gemini, GPT, DeepSeek, Llama, Qwen, and other open-source models, meaning they are not restricted to a single provider.
Awesome LLM Apps' Apache-2.0 license permits commercial use, and the README says directly: clone it, ship it, sell it.
The Awesome LLM Apps README says new templates are added weekly, distributed through the Unwind AI newsletter.
Awesome LLM Apps includes an AI Agent Framework Crash Courses section, such as a Google ADK crash course covering tools, memory, callbacks, and multi-agent patterns.
The problem it solves
Developers experimenting with LLM agents often start from a blank file and rebuild the same scaffolding, API client setup, tool wiring, a retrieval step, before they can test their actual idea. Awesome LLM Apps solves that by providing pre-built, runnable references for 100+ narrow use cases, from a travel-itinerary agent to a hash-chained audit trail for a trust-gated agent team, so that scaffolding work is already written and reviewable in the source.
Who should try it — and who should skip
Reach for Awesome LLM Apps if you learn by running and modifying working code: cloning the AI travel agent or a RAG tutorial gets you a functioning baseline in about 30 seconds, per the README's own timing. Skip it if you want one coherent framework with a stable API surface. This is 100+ independent scripts with separate dependencies, not a library, so you reconcile API keys and requirements across examples instead of following one onboarding path.