Awesome LLM Apps

What is in the repository

Awesome LLM Apps, maintained by Shubham Saboo, is less a list of links and more a collection of small projects that live in the repository itself. As of September 2026 it has about 139k GitHub stars and 20.5k forks, is released under the Apache 2.0 licence, and is updated frequently. Its top-level folders show the range:

• Starter AI agents and advanced AI agents, from single-purpose assistants to multi-agent teams.
• RAG tutorials, retrieval-augmented generation apps over documents, websites and databases.
• Advanced LLM apps, including chat with your data and memory-based apps.
• MCP AI agents, built on the Model Context Protocol for connecting models to tools.
• Voice AI agents and generative UI agents.
• Agent skills, always-on agents and an AI agent framework crash course.

Most apps are Python, often with a Streamlit interface, and use hosted models from OpenAI, Anthropic or Google, or local models through tools like Ollama.

Why complete applications teach better

Tutorials show a clean path through one feature. Applications show the parts nobody writes tutorials about: configuration, prompts that actually work, handling output that does not parse, and wiring retrieval, tools and a UI together. Those are the parts you need when you build your own.

Before you start

• Python basics and comfort with virtual environments and pip.
• An API key for at least one model provider, or a local model setup. Running many apps costs a little in API usage, so set a spending limit.
• Basic idea of embeddings and RAG helps for the retrieval folders.

How to use it well

1. Pick one app close to what you want to build, not the most impressive one.
2. Run it unchanged and confirm it works.
3. Read every file and sketch the data flow: input, prompt, model call, tools, output.
4. Change something real: swap the model, point it at your own documents, add an evaluation step.
5. Break it on purpose and fix it, then rebuild the core yourself without copying.

That last step turns a cloned demo into a portfolio project you can explain.

Where it falls short

Quality varies between entries, as with any large collection. Many apps are demos: little error handling, no tests, no evaluation and no thought for cost or security. Library versions move quickly, so some older apps need small fixes to run. Treat the code as a starting point, not a production reference.

Common questions

Is it free? Yes, the code is open source. You pay only for model API usage, if any.

Can I put a cloned app on my CV? Only after changing it substantially and understanding every part. Recruiters and interviewers can tell a fork from your own work.

Which folder should a beginner start with? Starter AI agents or the simpler RAG tutorials.

Best for: engineers who learn by dismantling working examples and want a head start on an LLM project.