Awesome LLM Apps: 100+ Runnable AI Agents and RAG Apps

Awesome LLM Apps is the rare “awesome list” that’s actually a folder full of things you can run. Skip the prompt-snippet sites. This is 100+ working AI agents and RAG apps, each one cloned, customized, and shipped end-to-end under an Apache-2.0 license.

View Awesome LLM Apps on GitHub

100+ apps, not 100 prompts

The “awesome” prefix usually signals another pile of links to skim and forget. This repo isn’t that. Every entry is a real codebase (a travel agent, a medical imaging diagnostician, a meme generator that drives a real browser, a breakup recovery chat) that you can clone, run with one command, and modify without a PhD in the underlying framework.

The through-line is what the maintainer calls “hand-built, tested end-to-end”. That’s not marketing. Each app has its own folder, requirements, and entry point. You can open the repo in your editor, pick the one that matches your problem, and have something demoable running in the time it takes to read its README.

Skills for Claude Code, Cursor, and Codex

The newest layer is the one most worth flagging if you live inside an agentic coding tool. “Agent skills” ship as installable packages (npx skills add <url>) that give Claude Code, Codex, Cursor, or any compatible agent new abilities. One command, plain English to use after that, real code underneath.

Standouts include Project Graveyard (autopsies your abandoned side projects and recommends which one to revive), an Advisor-Orchestrator-Worker loop that chains different models into different roles, and Self-Improving Agent Skills that rewrite themselves against eval scores. Each skill passes a security and eval CI gate before it lands, which is more than most prompt marketplaces can say.

From a 30-second clone to an always-on worker

The repo is organized by ambition. Starter agents are single-file apps (podcast generator, music generator, CSV analyst) that run with just an API key. Advanced agents go multi-step: a home renovation assistant that takes a photo and returns a photoreal redesign, an insurance claim team that resolves voice claims live, a fraud investigator that cross-examines public records.

The always-on tier is where it gets genuinely interesting for indie builders. These are long-running agents (a Hacker News briefing bot that reads while you sleep, a self-improving skill loop, a project graveyard scanner) designed to live on a server rather than in a terminal session. They’re closer to products than demos.

What to know before you treat it as a library

Two honest caveats. First, the apps lean on commercial APIs (OpenAI, Anthropic, Google) for most heavy lifting, so “free and open source” refers to the code, not the runtime bill. Budget accordingly. Second, quality varies across 100+ entries. The showcase agents are clearly more polished than the long tail, so start with the gallery images in the README and work outward from there.

The companion site, Unwind AI, publishes weekly tutorials that walk through building each new template. If a particular agent catches your eye but the README is too thin, the Unwind post usually fills in the why and how.

Run one now

Fastest path to give your coding agent a new skill:

npx skills add https://github.com/Shubhamsaboo/awesome-llm-apps/tree/main/agent_skills/project-graveyard

Or clone a starter app and run it directly:

git clone https://github.com/Shubhamsaboo/awesome-llm-apps.git
cd awesome-llm-apps/starter_ai_agents/ai_travel_agent
pip install -r requirements.txt
streamlit run travel_agent.py

Awesome LLM Apps earns its stars. If you’ve been collecting AI tutorials without shipping anything, pick one app from the gallery, get it running tonight, and modify one thing about it. That’s the entire point of the repo.