Awesome LLM Apps: 100+ Open-Source AI Agents You Can Clone, Customize, and Ship Today

Q: What is the difference between Starter, Advanced, and Multi-Agent categories?

Starter agents are single-file bots that demonstrate core patterns with minimal setup. Advanced agents add tools, memory, and multi-step reasoning for production use. Multi-agent teams involve multiple specialized agents collaborating on complex tasks through orchestration patterns.

Discover the most comprehensive open-source collection of AI agent and RAG applications — from starter bots to multi-agent teams — all ready to run with Claude, GPT, Gemini, and open-source models.

CoddyKit TeamJul 14, 2026 · 8 min read · 1,556 words

Awesome LLM Apps: 100+ Open-Source AI Agents You Can Clone, Customize, and Ship Today

Quick Answer: Awesome LLM Apps is the largest open-source collection of AI agent and RAG applications, featuring 100+ ready-to-run projects. Every app works with major LLMs (Claude, GPT, Gemini, DeepSeek, Llama, Qwen), ships under the Apache-2.0 license, and can be cloned and running in under 30 seconds. It covers everything from single-file starter agents to multi-agent teams with tools, memory, and voice capabilities.

If you've been following the AI developer ecosystem, you've probably noticed a shift: the conversation has moved from "how do I use an LLM?" to "how do I build a production-ready AI agent?". The Awesome LLM Apps repository by Shubham Saboo sits right at the center of this shift — and it's one of the most-starred AI collections on GitHub.

This isn't just a bookmark list. Every project in the repository is hand-built, tested end-to-end, and open-source. You can clone it, customize it, and ship it — whether you're building a weekend prototype or a production system.

The repository covers the full spectrum of modern AI agent development:

Starter Agents — Single-file bots that need just an API key
Advanced Agents — Production-style agents with tools, memory, and multi-step reasoning
Multi-Agent Teams — Multiple agents collaborating on complex tasks
Voice AI Agents — Speech-in, speech-out agents using real-time voice APIs
Generative UI Agents — Agents that render interactive UI components, not just text
Agent Skills — Modular capabilities you can install into coding agents
Always-On Agents — Background agents that run on schedules or events

Getting Started: Run Your First Agent in 30 Seconds

One of the best things about this collection is how frictionless it is to get started. Here's all you need to run a travel planning agent:

# Clone the repository
git clone https://github.com/Shubhamsaboo/awesome-llm-apps.git
cd awesome-llm-apps/starter_ai_agents/ai_travel_agent

# Install dependencies
pip install -r requirements.txt

# Run the agent
streamlit run travel_agent.py

That's it. You now have a personalized travel itinerary agent running locally. Swap in your API key for OpenAI, Anthropic, or Google, and you're ready to go.

Add Agent Skills to Your Coding Agent

The repository also introduces a new concept: Agent Skills — modular capabilities you can install into your coding agent with a single command:

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

This installs the "Project Graveyard" skill, which finds every side project you've abandoned, analyzes why each one died, and helps you finish the one worth going back to. It works with Claude Code, Codex, Cursor, and other coding agents.

Exploring the Agent Categories

🚀 Starter AI Agents

These are single-file agents designed as learning starting points. Each one runs with just an API key and demonstrates core agent patterns:

AI Blog to Podcast Agent — Turn any blog URL into a narrated podcast episode
AI Data Analysis Agent — Ask questions of any CSV or Excel file in plain English
AI Medical Imaging Agent — Diagnostic analysis of X-rays and scans with Gemini
AI Meme Generator — Makes memes by driving a real browser, not an image API
Mixture of Agents — Multiple LLMs answer, one aggregates the best response
xAI Finance Agent — Real-time stock analysis powered by Grok

🏗️ Advanced AI Agents

Production-style agents with tools, memory, and multi-step reasoning. These demonstrate real-world patterns for enterprise applications:

AI Deep Research Agent — Comprehensive web research with OpenAI Agents SDK and Firecrawl
AI System Architect Agent — Architecture reviews using DeepSeek R1 reasoning plus Claude
AI Self-Evolving Agent — Agents that rewrite their own workflows with EvoAgentX
AI Fraud Investigation Agent — Cross-references public records to flag anomalies
Trust-Gated Multi-Agent Research Team — Every agent verified, every action in a hash-chained audit trail

👥 Multi-Agent Teams

Perhaps the most exciting category. These projects demonstrate multiple agents collaborating on complex, cross-domain tasks:

AI VC Due Diligence Agent Team — Multi-agent startup investment analysis with Gemini
AI Legal Agent Team — Research, contract analysis, and strategy from a full legal bench
AI Recruitment Agent Team — Resume screening to interview scheduling, end to end
Multimodal Coding Agent Team — Snap a photo of a coding problem, get a sandboxed solution
AI Competitor Intelligence Team — Structured competitor teardowns from their own websites

# Example: AI Finance Agent Team in 20 lines of Python
from phi.agent import Agent
from phi.model.openai import OpenAIChat
from phi.tools.yfinance import YFinanceTools

analyst = Agent(
    name="Analyst",
    model=OpenAIChat(id="gpt-4o"),
    tools=[YFinanceTools(stock_price=True, analyst_recommendations=True)],
    instructions=["Analyze stock data and provide insights"],
)

reporter = Agent(
    name="Reporter",
    model=OpenAIChat(id="gpt-4o"),
    instructions=["Compile analyst insights into a report"],
)

team = Agent(
    team=[analyst, reporter],
    instructions=["First analyze, then compile a report"],
    show_tool_calls=True,
)

team.print_response("Compare AAPL and MSFT for Q2 2026")

🎙️ Voice AI Agents

Speech-in, speech-out agents using real-time voice APIs. These are perfect for customer-facing applications:

AI Audio Tour Agent — Self-guided audio tours from your location and interests
Customer Support Voice Agent — Voice answers grounded in your own docs
Insurance Claim Live Agent Team — Real-time voice claim intake with Gemini Live
Voice RAG Agent — Ask your PDFs questions, hear the answers

🎛️ Generative UI Agents

Agents that don't just output text — they render interactive UI components like forms, cards, charts, and editable plans:

Generative UI Starter — A chat-driven kanban board you and the agent work on together
AI Dashboard Canvas Agent — Describe a dashboard in chat, charts assemble on a live canvas
AI MCP App Builder — Describe an MCP app, get a live sandboxed instance back
AI Shadcn Component Generator — Chat your way to production-ready shadcn components

Real-World Example: Building a Customer Support Agent

Let's say you want to build a customer support agent for your SaaS product. Here's how you'd use the Awesome LLM Apps patterns:

from phi.agent import Agent
from phi.model.anthropic import Claude
from phi.tools.duckduckgo import DuckDuckGo
from phi.knowledge.pdf import PDFUrlKnowledgeBase
from phi.vectordb.pgvector import PgVector2

# Step 1: Set up knowledge base from your docs
knowledge_base = PDFUrlKnowledgeBase(
    urls=["https://yourcompany.com/docs/support-guide.pdf"],
    vector_db=PgVector2(
        db_url="postgresql://user:pass@localhost:5432/support_db",
        collection="support_docs"
    )
)
knowledge_base.load()

# Step 2: Create the support agent
support_agent = Agent(
    model=Claude(id="claude-sonnet-4-20250514"),
    knowledge=knowledge_base,
    tools=[DuckDuckGo()],
    instructions=[
        "Answer customer questions using the knowledge base first.",
        "If the answer isn't in the docs, search the web.",
        "Always be empathetic and professional.",
        "If you can't resolve, suggest escalating to a human agent."
    ],
    show_tool_calls=True,
    markdown=True
)

# Step 3: Run it
support_agent.print_response(
    "How do I reset my API key? I'm getting a 401 error."
)

This pattern — knowledge base + web search + structured instructions — is one of the most commonly used in the repository. You can adapt it for any domain: legal, medical, financial, or technical support.

Key Benefits of Using Awesome LLM Apps

🆓 100% Free & Open Source — Apache-2.0 license means you can use it commercially without restrictions
⚡ Instant Setup — Most agents run in under 30 seconds with just a git clone and pip install
🔌 Multi-Model Support — Works with Claude, GPT, Gemini, DeepSeek, Llama, Qwen, and open-source models
📚 Learning Path — Progress from simple starter agents to complex multi-agent systems
🏭 Production-Ready Patterns — Real-world patterns with tools, memory, vector databases, and audit trails
🔒 Security-First — Skills ship with security + eval CI gates; trust-gated agent teams for sensitive work
🗣️ Voice-Ready — Built-in support for real-time voice APIs (Gemini Live, OpenAI Realtime)
🔄 Self-Improving — Some agents can rewrite their own workflows using EvoAgentX
📬 Weekly Updates — New templates and agent patterns added regularly

FAQ

What is Awesome LLM Apps?

Awesome LLM Apps is an open-source GitHub repository containing 100+ hand-built, tested AI agent and RAG applications. It covers everything from single-file starter agents to complex multi-agent teams, all available under the Apache-2.0 license.

Which LLM providers does it support?

The repository supports all major LLM providers including OpenAI (GPT-4o, o1), Anthropic (Claude), Google (Gemini), xAI (Grok), DeepSeek, and open-source models like Llama and Qwen. Most agents let you swap models with a single line change.

Can I use these agents in production?

Yes. The advanced agents and multi-agent teams are designed with production patterns including tools, memory, vector databases, audit trails, and trust gates. The Apache-2.0 license also allows commercial use without restrictions.

How do Agent Skills work?

Agent Skills are modular capabilities you can install into coding agents (Claude Code, Codex, Cursor) with a single npx command. They add specific abilities to your agent — like analyzing abandoned projects or orchestrating multi-model workflows — without modifying the agent's core configuration.

What's the difference between Starter, Advanced, and Multi-Agent categories?

Starter agents are single-file bots that demonstrate core patterns with minimal setup. Advanced agents add tools, memory, and multi-step reasoning for production use. Multi-agent teams involve multiple specialized agents collaborating on complex tasks through orchestration patterns.

Do I need to pay for API keys?

Most agents require API keys from providers like OpenAI, Anthropic, or Google, which have usage-based pricing. However, many agents also support open-source models (Llama, Qwen, DeepSeek) that you can run locally for free using Ollama or similar tools.

How often are new agents added?

New templates and agent patterns are added weekly. You can follow the repository on GitHub or subscribe to the Unwind AI newsletter to get updates delivered to your inbox.

Can I contribute my own agents?

Yes, the repository accepts contributions. You can submit a pull request with your agent following the project's structure and testing guidelines. All contributions must be under the Apache-2.0 license.

ProgrammingTutorialCoddyKit