Don't Build AI Apps From Scratch. Use This Free Repo Instead

Before you build your next AI app from zero, check whether someone has already built it. There’s a good chance they have, and that the code is sitting in one GitHub repo.

Awesome LLM Apps by Shubham Saboo has more than a hundred working AI apps in a single repository: agents, voice assistants, agent teams, generative UI, MCP agents and more. It has over 144,000 stars and it’s completely free. In the video I go through every section and then run one app on my own computer. This post covers the same ground in written form, with every command you need.

🎥 Watch the Video

👉 Watch on YouTube: Don’t Build AI Apps From Scratch

What Awesome LLM Apps Is

Many “awesome” repos on GitHub are just lists of links. This one is different: every app lives inside the repo. You open a folder and the code is right there.

• License: Apache 2.0. You can clone it, change it, use it in your own projects and even build a product on top of it.
• Models: it isn’t locked to one AI company. Apps use Claude, GPT, Gemini, DeepSeek, Llama, Qwen and other open models.
• Local options: some apps run fully on your own machine with Ollama, with no API key at all.

🔗 Repo: github.com/Shubhamsaboo/awesome-llm-apps

The README splits everything into 16 sections. Here’s what’s in each one.

Agent Skills

These aren’t browser apps. They’re skills you add to a coding agent like Claude Code, Cursor or Codex, and then use in plain English.

• Project Graveyard finds the side projects you’ve abandoned, explains why each one died, and helps you finish the one worth going back to.
• First Reader simulates real readers going through your draft and tells you where they get bored or stop, without rewriting a word.
• Scope Creep Detector checks whether a code change grew bigger than it was meant to be.
• Commit Archaeologist digs through Git history to explain why a piece of code exists.
• Dependency Doctor flags old, broken or risky packages in your dependency file.
• Thinking Out Loud turns a messy voice ramble into a clean, short brief.
• Self-Improving Agent Skills is a project where skills test themselves and get better automatically.

Standalone Apps and Tools

• Excaliflow: hand-drawn system diagrams with animated arrows, exportable as video.
• Needle: a Chrome extension that searches a page by meaning, not exact words.
• Ripple: watches a Google Doc and finds what else must change when you edit one thing.

Starter AI Agents

If you’re new to agents, start here. They’re small and easy to read.

Advanced AI Agents

Once you’re comfortable, this section gets serious:

• Home Renovation Agent: photos of your room in, renovation plan and realistic images out.
• Deep Research Agent: full web research and a written report on any topic.
• System Architect Agent: reviews software architecture with a reasoning model plus Claude.
• Investment Agent: compares stocks and builds a report from Yahoo Finance data.
• Earnings Call Analyst: turns YouTube earnings calls into a synced analysis workspace.
• Movie Production Agent: one-line idea in, script drafts and casting ideas out.
• Journalist Agent: researches, writes and edits articles.
• Health & Fitness Agent: diet and workout plans based on your goals.
• Meeting Agent: context, industry insights and a strategy brief before a meeting.
• Fraud Investigation Agent: cross-checks public records to flag inconsistencies.
• Self-Evolving Agent: rewrites its own workflow to improve over time.

Always-on Agents

These run on a schedule instead of waiting for you to ask:

• Hacker News Briefing Agent: a ranked daily brief sent to Slack or email.
• Release Radar Agent: watches your dependencies for breaking changes, security fixes and major updates.

Multi-agent Teams

Instead of one AI doing everything, a team of agents each takes one job:

• Finance Agent Team: a financial analyst team in about twenty lines of Python.
• Legal Agent Team: research, contract analysis and strategy, with a local version too.
• Recruitment Agent Team: from resume screening to interview scheduling.
• Real Estate Agent Team: property search, market analysis and recommendations.
• Teaching Agent Team: a complete learning path from a “faculty” of agents.
• Game Design Agent Team: full game concepts from design specialists.
• Multimodal Coding Agent Team: photo of a coding problem in, working solution out.
• UI/UX Feedback Team: reviews your landing page and generates an improved version.
• Competitor Intelligence Team: a competitor breakdown built from their own websites.
• LLM Panel Team: three models review the same code change blind, then argue it out.

Voice AI Agents

• Audio Tour Agent: a self-guided audio tour based on your location, interests and pace.
• Customer Support Voice Agent: answers out loud using your own documents.
• Insurance Claim Live Agent Team: takes your claim by voice, looks at the damage through your webcam and sketches what happened.
• Voice dictation agent: an open-source agent that types wherever you talk.

Generative UI and Agentic Frontends

Here the AI builds the interface, not just the reply:

• Dashboard Canvas Agent: describe a dashboard in chat and the charts appear on a live canvas.
• Shadcn Component Generator: chat your way to ready-to-use UI components.
• MCP App Builder: describe an app, get a live working version.
• Financial Coach: your budget and savings plan as interactive cards.
• Generative UI Starter: a kanban board you and the AI work on together.

Game-Playing Agents

• Chess Agent: Agent White vs Agent Black, with every move validated.
• Tic-Tac-Toe Agent: two different AI models play each other.
• 3D PyGame Agent: writes game code and runs it live in the browser.

MCP AI Agents

MCP (Model Context Protocol) is the standard way to connect AI to outside tools and data.

• Browser MCP Agent: control a real browser by telling it what to do.
• GitHub MCP Agent: explore any repo in plain English.
• Notion MCP Agent: talk to your Notion pages from the terminal.
• Travel Planner MCP Agent: trips built from live Airbnb and Google Maps data.
• Dependency Upgrade Advisor: tells you what will break before you upgrade.

RAG, Memory and Chat with X

• RAG: more than twenty apps that answer questions from your own documents, including fully local versions.
• LLM Apps with Memory: apps that remember you, like a travel agent that learns your preferences and a local ChatGPT clone with per-user memory.
• Chat with X: chat with your Gmail inbox, a PDF, a GitHub repo, YouTube videos, research papers or a Substack archive. Most are around thirty lines of code.

Optimization, Fine-tuning and Crash Courses

• Optimization tools that cut the tokens you send, lowering your API bill.
• Fine-tuning examples for small open models like Gemma and Llama, one of which runs free on Google Colab.
• Crash courses for Google’s Agent Development Kit and the OpenAI Agents SDK.

Before You Run Any App

Open the app’s folder and read two files: the README and requirements.txt.

• If you see “OpenAI API key” (or another provider’s key), you’ll pay for that model.
• If you see Ollama, it usually runs locally on your own computer for free.

Demo: Run the Resume & Job Matcher Locally

I picked the Resume & Job Matcher from advanced_llm_apps/resume_job_matcher. You give it a resume PDF and a job description, and it returns a fit score out of 100, your strengths and what to improve. It runs on Ollama, so there’s no API key.

You need: Git, Python 3 and Ollama installed.

1. Download only this app

The repo is big, so use a sparse checkout to download just one folder:

1
2
3
4

git clone --depth 1 --filter=blob:none --sparse https://github.com/Shubhamsaboo/awesome-llm-apps.git
cd awesome-llm-apps
git sparse-checkout set advanced_llm_apps/resume_job_matcher
cd advanced_llm_apps/resume_job_matcher

2. Create a virtual environment and install packages

1
2
3
4

python -m venv .venv
.venv\Scripts\activate          # Windows
# source .venv/bin/activate     # macOS / Linux
pip install -r requirements.txt

3. Get the model

Open app.py. On line 68 the model is set to "llama3". Either pull that model, or change the word to a model you already have in Ollama:

1

ollama pull llama3

4. Start the app

1

streamlit run app.py

The app opens in your browser. Upload the resume, paste or upload the job description, and in a few seconds you get a fit score, your strengths and what to improve for that exact job. Everything runs on your own computer: no account, no API key and no cost.

If you get an error

It’s almost always because Ollama isn’t running. Open the Ollama app, or run this in another terminal and try again:

1

ollama serve

Takeaways

Awesome LLM Apps is more than a hundred AI apps across sixteen sections, from simple starter agents to voice agents, agent teams and full crash courses. All of it is free and open source.

If you’re learning AI, don’t just read about it. Pick one app, run it, then change it and make it your own.

Which app are you trying first? Tell me in the comments on the video. Thanks for reading, and see you in the next one.