Build an AI App From a Free Working Example

Awesome LLM Apps is a free GitHub collection of more than 100 AI apps that already work. Pick the one closest to your idea, run it as-is, then have Claude Code change it one step at a time.

What this helps you do

Starting from a blank page is the slowest way to build your first AI app. Every project in this collection has its own folder, its own code, and its own setup steps. You borrow a working app, prove it runs on your computer, and then shape it into the thing you want.

This page gives you eight good starters with their exact folders, the setup you need, and two prompts for Claude Code: one to run the original, one to make a single change.

Before you start

• Claude Code installed on your computer. The Claude Code docs walk you through it. New to Claude? Start with the Claude Setup Checklist.
• Git and Python 3.8 or newer. Most starters run in Python. The dashboard starter also needs Node.js 18 or newer.
• An API key for most starters. An API key is a password that lets the app use an AI model and bills your account for what it uses. Each starter's README names the keys it needs. One starter below runs with no key at all.
• One small job in mind, such as "answer questions from my event schedule" or "turn my blog post into audio."

Terms to know

• Repo: short for repository, a project folder stored on GitHub.
• Clone: download a full copy of a repo to your computer.
• README: the setup instructions inside each folder. Read it before you run anything.
• Streamlit: a tool many of these apps use to show a simple web page on your own computer, usually at http://localhost:8501.
• Starter: the original app, before you change anything.

Step-by-step

1. Open a terminal, go to the folder where you keep projects, and clone the collection:

git clone https://github.com/Shubhamsaboo/awesome-llm-apps.git

2. Pick one starter from the list below. Choose the one closest to what you want, not the most impressive one.
3. Open that starter's folder in Claude Code and read its README together. Note which API keys it needs.
4. Paste the first prompt from Copy this. Claude Code installs what the app needs and starts it. Approve each install before it runs.
5. Use the original app once, exactly as built. You should see its page open in your browser and give you a real result. If it doesn't run yet, fix that before changing anything.
6. Now paste the second prompt with one change, such as a new input, a new data source, or a new design. Run the app again and check the result.
7. Repeat with the next single change. If something breaks, you know which change did it.

Eight good starters

All folder paths start inside the awesome-llm-apps folder you cloned.

• Blog to podcast. Paste a blog link, get a narrated audio episode. Folder: starter_ai_agents/ai_blog_to_podcast_agent. Needs OpenAI, ElevenLabs, and Firecrawl keys.
• Room renovation planner. Upload a photo of your room and get a renovation plan, a budget, and new images of the redesigned space. Folder: advanced_ai_agents/multi_agent_apps/ai_home_renovation_agent. Needs a Google Gemini key. This one runs in a Google tool called ADK Web, so follow its README closely.
• Meme maker. Describe a meme and an agent opens a real web browser, picks a template, and writes the captions. Folder: starter_ai_agents/ai_meme_generator_agent_browseruse. Needs one key from Anthropic, OpenAI, or DeepSeek.
• Dashboard you can change. Describe a dashboard in chat and charts appear on a live canvas. Folder: generative_ui_agents/ai-dashboard-canvas-agent. Needs Node.js and a Google key.
• Research agent. Give it a topic and a small team of agents searches the web and writes a cited report. Folder: starter_ai_agents/openai_research_agent. Needs an OpenAI key.
• Chat with a PDF. Upload a document and ask it questions. Folder: advanced_llm_apps/chat_with_X_tutorials/chat_with_pdf. Needs an OpenAI key.
• Local document agent, no API key. Answers questions from your documents with everything running on your computer. Folder: rag_tutorials/local_rag_agent. Needs Ollama and Docker installed instead of a key.
• Travel planner. Builds a day-by-day itinerary you can download to your calendar. Folder: starter_ai_agents/ai_travel_agent. Needs OpenAI and SerpAPI keys, or use its local version with Ollama.

Copy this

Paste this into Claude Code with the starter's folder open. It runs the original before anything changes.

This folder is a working AI app from the awesome-llm-apps collection. Read the README and the code, then help me run the ORIGINAL app exactly as written. Do not change any app code yet.
1. Tell me in plain language what the app does and which API keys it needs.
2. Set up a virtual environment and install the requirements. Ask before installing anything outside this folder.
3. Put my keys in a .env file that stays out of Git. Ask me to paste each key myself.
4. Start the app and tell me the address to open in my browser.
5. If it fails, show me the exact error and fix only the setup, not the app.

Once the original works, paste this. Change only one thing per round.

The original app runs. Make exactly ONE change: [YOUR CHANGE, e.g. "answer questions from my event schedule PDF instead of the sample file"].
Before editing, tell me which files you will touch and why. Keep everything else the same. After the change, run the app again and tell me how to test it. If it breaks, undo this change so we are back to the working version.

Things to know

• API calls cost money. Most starters call a paid model each time you use them. Check each provider's pricing page and set a spending limit before you test.
• Keep keys out of your code. Store them in a .env file and never share it or upload it to GitHub.
• Use the folder paths on this page. If a cd command in a README fails, open the folder listed above instead. Claude Code can find it for you.
• You can build on them. The collection is released under the Apache-2.0 license, which allows personal and commercial use.
• Change order matters. Change the input first, then the logic, then the design. One change per round keeps every break easy to trace.

If something goes wrong

• "Module not found" error: the requirements didn't install in the environment the app uses. Ask Claude Code to reinstall them in the project's virtual environment.
• The app opens but every answer fails: a key is missing or wrong. Check the key names in the README match your .env file.
• The browser page never opens: look in the terminal for the address, often http://localhost:8501, and open it yourself.
• It worked, then broke after a change: ask Claude Code to undo the last change, confirm the original runs again, then try a smaller change.
• The local agent can't connect: make sure Ollama and Docker are both running before you start it.

Final check

You pass when the original starter runs on your computer and gives a real result, and your changed version still runs after each single change you made.