awesome-ai-apps 收录 137 个可运行 LLM 应用与 Agent 项目

137 runnable LLM apps and agents, from 5-minute framework starters to production-grade multi-agent systems.

Browse projects · Quick start · Contribute · Sponsor · Newsletter

Awesome AI Apps is a hands-on collection of projects, tutorials and recipes for building LLM-powered applications: text agents, voice assistants, RAG apps, MCP tools and fine-tuning pipelines. Each project lives in its own folder with a README, its own dependencies and a .env.example, so you can clone the repo, open one folder and run it.

💎 Sponsors

A huge thank you to the companies supporting Awesome AI Apps.

Want your product in front of developers building with LLMs?
Become a sponsor →

🧭 Browse by Category

Also here: Tutorials & Videos · Quick Start · Contributing · Contributors · Maintainers · License

🚀 Featured AI Apps

🧩 Starter Agents

Quick-start agents for learning and extending different AI frameworks. 21 projects

• AutoGen Tool-Calling Starter: Microsoft AutoGen AssistantAgent with a custom tool, powered by Nebius Token Factory
• AWS Strands Agent Starter: Weather report agent using AWS Strands SDK
• CAMEL AI Model Benchmark: Performance benchmarking tool comparing various AI models
• Coding Harness Starter: OpenAI Agents SDK coding loop with plan, approval-gated edits, fixed tests, and bounded review, powered by Nebius Token Factory
• CrewAI Research Crew: Multi-agent research team example
• Docker cagent Multi-Agent Starter: Open-source customizable multi-agent runtime by Docker
• DSPy Optimization Starter: DSPy framework for building and optimizing AI systems
• Google Agent Development Kit Starter: Google Agent Development Kit starter template
• Hacker News Trend Analyst (Agno): Agno-based agent for trend analysis on Hacker News
• Hugging Face smolagents Starter: Hugging Face smolagents code-first web-search agent
• KAOS Kubernetes Multi-Agent Starter: Kubernetes-native multi-agent system with MCP tools and in-cluster LLM
• LangChain Tool-Calling Starter: LangChain tool-calling agent with create_tool_calling_agent + AgentExecutor, powered by Nebius
• LangGraph ReAct Agent Starter: LangGraph prebuilt ReAct agent (create_react_agent) with custom tools, powered by Nebius
• Letta Stateful Memory Agent: Stateful agent with persistent long-term memory across sessions
• LlamaIndex Task Manager: LlamaIndex-powered task assistant
• Mastra Tool-Calling Starter: TypeScript-first agent with a custom tool powered by Nebius Token Factory
• Microsoft Agent Framework Starter: Multi-agent travel planning demos built on Microsoft Agent Framework
• OpenAI Agents SDK Starter: OpenAI Agents SDK with email helper and haiku writer examples
• PydanticAI Weather Bot: Real-time weather information agent
• Sayna Realtime Voice Agent: Real-time voice infrastructure with multi-provider STT/TTS (Deepgram, ElevenLabs, Azure, Google) and WebSocket streaming
• Semantic Kernel Starter: Microsoft Semantic Kernel ChatCompletionAgent with plugin-based tool calling

🪶 Simple Agents

Straightforward, practical use-cases for everyday AI applications. 20 projects

• Agno Agent Examples: Simple to multi-agent examples with web search and a knowledge base
• Agno Agent UI: Interactive UI for web and finance agents
• AI Agent Registry Explorer: Find and compare AI agents across NANDA, MCP, Virtuals, A2A, and ERC-8004 registries
• AI Expense Tracker: Chat-based expense tracking with Nebius Token Factory and a Postgres backend that agents provision on first run
• Calendar Assistant: Calendar scheduling integration with Cal.com
• Cost-Aware Model Router (RouteLLM): Intelligent model routing with RouteLLM (GPT-4o-mini vs Nebius Llama) for cost optimization
• Email-to-Calendar Assistant: AI-powered Gmail reader and Google Calendar manager
• Financial Reasoning Agent: Step-by-step financial reasoning demonstration
• Human-in-the-Loop Agent: HITL actions for safe AI task execution
• LangChain Operations Agent Collection: Nebius-powered incident response, support, vendor risk, and data quality agents with typed outputs and guarded tools
• Macro Event Brief Agent: Scheduled macro releases with last prints, publication times and sources, plus an optional Nebius brief that cites each record
• Mastra Weather Bot: Weather updates using Mastra AI framework
• Natural-Language Database Assistant: Natural language database queries with GibsonAI and LangChain
• Natural-Language SQL Agent (LangChain): Natural-language-to-SQL data agent with LangGraph, Nebius, read-only SQL safety, and Streamlit charts
• Nebius Chat: Chat interface for Nebius Token Factory
• Newsletter Generator: AI-powered newsletter builder with Firecrawl integration
• Stock Market Finance Agent: Real-time stock and market data tracking agent
• Stock Portfolio Analyst: Live portfolio valuation, concentration analysis, risk flags, and rebalancing ideas with Agno
• VoyageCompass Travel Planner: LangChain and Nebius travel planner with weather, research, currency conversion, budgets, and packing tools
• Web Automation Agent: Browser automation agent using Nebius and browser-use

🎙️ Voice Agents

Real-time voice assistants and streaming speech pipelines, including LiveKit, Pipecat, Gradium, and VoxCode (Deepgram + Cursor SDK). 9 projects

• AI Pitch Coach (Gradium + Nebius): Conversational pitch coach using Gradium STT/TTS, LangChain orchestration, and Nebius reasoning
• Customer Support Voice Agent (LiveKit): Nebius-powered voice support agent with context-preserving AI manager handoff, noise cancellation, and inactivity handling
• Gemini Realtime Voice Agent (LiveKit): LiveKit Agents with Google Gemini Live (gemini multimodal realtime) for low-latency voice conversations in a LiveKit room
• Healthcare Voice Contact Center: Pipecat healthcare contact center with appointment booking, FAQ handling, and supervisor escalation
• Multilingual Voice Agent (Pipecat + Sarvam): Pipecat voice pipeline with Sarvam STT/TTS and OpenAI for chat; WebRTC (browser) or Daily transport via the Pipecat runner
• RSVP Confirmation Voice Agent (LiveKit): Outbound voice agent that calls attendees, confirms RSVPs, and updates a JSON-backed event database
• Speed-to-Lead Sales Voice Agent: LiveKit-based voice agent that calls inbound leads instantly, routes them to specialists, and logs to a mock CRM
• VoxCode: Deepgram + Cursor Voice Coding Agent: Local voice workspace for codebase summaries and architecture Q&A; Deepgram Voice Agent orchestration, Nebius reasoning, and optional Cursor SDK file inspection and edits
• Web-Search Voice Agent (LiveKit): LiveKit + Gemini realtime voice agent with an Olostep-backed web_search tool for fresh, source-cited answers

🗂️ MCP Agents

Examples using Model Context Protocol for external tool integration. 14 projects

• Couchbase LangGraph MCP Agent: LangChain ReAct agent with Couchbase integration
• Couchbase MCP Server: Couchbase database integration with MCP protocol
• Custom MCP Server Starter: Custom MCP server implementation example
• Documentation Q&A MCP Agent: Documentation Q&A agent with MCP
• Documentation RAG MCP Server: Semantic RAG documentation and Q&A system
• GibsonAI Database MCP Agent: Conversational AI agent for managing GibsonAI database projects and schemas
• GitHub MCP Agent: Repository insights and analysis via MCP
• GitHub MCP Agent Starter: GitHub repository analyzer starter template
• Hotel Finder Agent: Hotel search and booking using MCP integration
• Sandboxed Code Execution MCP Agent (Docker + E2B): Secure AI agent for running agents in sandboxed Docker environments via MCP Gateway
• Secure Database MCP Agent (MCP Toolbox): Secure ecommerce agent over PostgreSQL and MongoDB; MCP Toolbox enforces per-user data access, least-privilege roles, and authorized tools
• Secure MCP Access Agent (ScaleKit + Exa): Security-focused MCP integration with Exa search
• Self-Healing Text-to-SQL Agent (Okahu): Self-healing Text-to-SQL demo using Okahu Cloud traces via hosted MCP
• Taskade MCP Agent: AI-powered workspace agent for managing projects, tasks, and workflows via Taskade MCP

🧠 Memory Agents

Agents with advanced memory capabilities for context retention and personalization. 13 projects

• AI Research Consultant with Long-Term Memory: AI-powered consulting agent using Memori v3 as a long-term memory fabric and ExaAI for research
• arXiv Researcher Agent with Memori: Research assistant using OpenAI Agents and GibsonAI Memori
• AWS Strands Persistent Memory Agent: AWS Strands agent enhanced with the Memori memory system
• Blog Writing Agent: Personalized blog writing agent with memory for style consistency
• Brand Reputation Monitor: AI-powered brand reputation monitoring with news analysis and sentiment tracking
• Customer Support Voice Agent: Voice-enabled customer support assistant with Memori v3 and Firecrawl for knowledge base management
• Engineering Content Agent: Chat-first Agno app that turns HN demand, DEV.to supply gaps, and Weaviate Engram memory into a developer trend digest plus DevRel talk and blog ideation via Nebius
• Job Search Agent: Job search agent with memory for preference tracking
• Persistent Memory Agent (Agno): Agno-based agent with persistent memory capabilities
• Product Launch Agent: Competitive intelligence tool for analyzing competitor product launches
• Social Media Agent: Social media automation agent with memory for brand voice
• Study Coach Agent: AI-powered study coach with Memori v3 and LangGraph for multi-step verification of understanding
• YouTube Trend Agent: YouTube channel analysis agent with Memori, Agno, and Exa for trend analysis and video ideas

📚 RAG Applications

Retrieval-augmented generation examples for document understanding and knowledge bases. 18 projects

• Agentic RAG with Agno and GPT-5: Agentic RAG implementation with Agno and GPT-5
• Agentic Typed RAG with LlamaIndex: Typed, citation-verified RAG with structured answers, local document parsing, and deterministic refusal for weak evidence
• Codebase Q&A RAG: Conversational code explorer and documentation assistant
• Enterprise Contextual RAG: Enterprise-level RAG with managed datastores and quality evaluation
• Gemma 3 Document OCR: OCR-based document and image processor using the Gemma 3 model
• GraphRAG with Neo4j: Knowledge graph extraction and Cypher-backed retrieval with Neo4j and Nebius
• LiteParse Invoice & Receipt Auditor: Local OCR with LiteParse bounding boxes, Nebius LLM audit for math errors and duplicate charges, evidence pinning on scans, and an LLM batch summary
• LlamaIndex RAG Starter: LlamaIndex and Nebius RAG starter template
• LLM and RAG Debugger (WFGY 16-Problem Map): 16-mode map-based debugger for LLM and RAG bugs
• Multi-PDF RAG Analyzer: Multi-PDF chat and analysis system
• Nebius RAG Starter: Basic RAG implementation with Nebius for quick starts
• NVIDIA Nemotron Document OCR: OCR-based document and image parsing using NVIDIA Nemotron-Nano-V2-12b
• Production PDF RAG with Reranking: Production-shaped PDF RAG with contextual retrieval, Qdrant hybrid search, reranking, streaming answers, upload ingestion, and clickable citations
• Qwen3 PDF RAG Chat: PDF chatbot interface built with Streamlit
• Resume Optimizer: AI-powered resume optimization and enhancement tool
• Trustworthy RAG: Citation verification, per-claim evidence checks, and hallucination scoring for RAG answers
• Video Q&A with Timestamp Citations: Multimodal video retrieval with Gemini embeddings, Weaviate, Nebius, and clickable timestamp citations
• Web-Augmented Agentic RAG: Advanced RAG with CrewAI, Qdrant, and Exa for hybrid search capabilities

🔬 Advanced Agents

Complex multi-agent pipelines for production-ready end-to-end workflows. 36 projects

• AI Hedge Fund Research Team: Agentic workflow for comprehensive financial analysis
• AI Trend Research Agent: AI trend mining and analysis with Google ADK
• Candidate Profile Analyzer (Candilyzer): Candidate analysis tool for GitHub and LinkedIn profiles
• Car Finder Agent: AI-powered used car recommendation system with CrewAI and MongoDB
• Clinical Early-Warning Agent: Nebius Token Factory tool-calling agent that spots patient deterioration with NEWS2, qSOFA, personal baselines and trends
• Conference Proposal Generator: Automated conference proposal generation system
• Conference Talk Abstract Generator: Automated talk abstract generation with Google ADK and Couchbase
• Contract Review Crew (Paralegal): CrewAI contract review workflow for clause extraction, risk analysis, and recommended redlines
• Cosmos Arena Debate Council: Multi-agent debate council built with LangGraph and the NVIDIA Cosmos reasoning model via Nebius Token Factory
• Customer Support Resolution Agent: LangChain and Nebius support agent with knowledge-base retrieval, order lookup, and human ticket escalation
• Deep Research + Writing Agents Workshop: Nebius-powered LangChain MCP workshop with Exa research, Gemini image generation, and Okahu/Monocle eval observability
• Deep Research Workflow: Multi-stage research agent with Agno and ScrapeGraph AI
• Due Diligence Agent: Multi-agent company due diligence pipeline with AG2 and TinyFish deep web scraping
• Financial Document OS: Turns financial PDFs into an editable relational database with source-level citations
• Financial Market Data Service: FastAPI server for stock data and predictions with Agno
• Financial Research Agent (AgentField): Financial research agent with AgentField
• GitHub and LinkedIn Job Finder: LinkedIn job search automation with Bright Data integration
• Jev Social Research Agent: Typed Jev routing, local socai CLI evidence, and evidence-linked Nebius synthesis for Instagram, TikTok, and LinkedIn research
• Local File-Editing Agent Prototype: Local coding-agent prototype with file discovery, reading, and editing tools
• Maintainer Intelligence Brief: Weekly open-source intelligence briefs from community, security, and document signals with source citations
• Meeting Assistant Agent: Automated meeting notes and task creation from conversations
• Multi-Agent Coding Harness: Deep LangGraph coding crew with planning, repository exploration, human-gated file edits, and E2B-sandboxed test loops
• Nebius Autonomous Pipeline Optimizer: NYC taxi analytics pipeline optimizer with iterative code search using real-time or batch Nebius Token Factory inference
• Pre-Meeting Intel Agent (Briefing Room): LangGraph plan-research-reflect loop with Tavily web search that turns a company name into a cited one-page meeting brief
• Price Monitoring Agent: Price monitoring and alerting agent powered by CrewAI, Twilio, and Nebius
• Prompt Format Benchmark: Benchmark harness for comparing XML, JSON, and Markdown prompt formats across accuracy, latency, and token usage
• Sandboxed Browser Game Generator: FastAPI multi-agent studio that generates sandboxed browser games from one prompt using Pydantic AI and GLM-5.2 on Nebius
• SEO Content Strategy Team: SEO content optimization workflow with Agno and SerpAPI for Google AI Search ranking
• Shark Tank Pitch Practice Agent: Interactive 3D Shark Tank room with three Nebius-backed Mastra investor sharks, round-based Q&A, deal memos, and SQLite report history
• Software Engineering Model Arena: Nebius Token Factory benchmark for two coding models with seven curated challenges, weighted local hidden tests, partial-credit scoring, and an independent judge
• Startup Go-to-Market Strategy Agent: Go-to-market strategy and competitive analysis agent
• Startup Idea Validator Agent: Agentic workflow to validate and analyze startup ideas
• Temporal Agents: Examples of Temporal-based AI agents
• Temporal Transaction Agent Evaluation (Okahu + Monocle): Evaluation and regression loop for a Temporal fraud-processing agent, using OpenTelemetry traces, deterministic checks, and LLM-graded Okahu evaluations
• Web Intelligence Agent: Mastra multi-agent pipeline that turns Olostep web evidence into Nemotron-verified case studies with SQLite persistence and a Velt audit trail
• Workflow Audit Trail (FlowSentinel): Next.js workflow command center with Nebius Nemotron reasoning, n8n orchestration, Velt activity logs, and optional Tailscale Funnel exposure

🧬 Fine-Tuning

End-to-end examples of fine-tuning open-source LLMs, from data prep to deployment. 6 projects

• Customer Support Fine-Tuning with Data Lab: Teacher-student distillation workflow to generate support data, curate it in Data Lab, fine-tune, and deploy
• Insurance Claims Fine-Tuning: Data Lab, LoRA fine-tuning, and a Gradio comparison app for insurance claims
• Legal Tech Fine-Tuning (Self-Hosted): Fine-tunes Gemma on UK legislation with LoRA, serves with vLLM, and exposes a FastAPI layer
• Legal Tech Fine-Tuning (Token Factory): Managed LoRA fine-tuning on Nebius Token Factory with private model deployment
• Open-Source LLM Fine-Tuning on Token Factory: Colab-first LoRA fine-tuning walkthrough to upload a dataset, train, monitor, and deploy
• Standalone Customer Support Fine-Tuning (Colab): Fully standalone Colab notebook for the customer-support distillation and fine-tuning flow

📺 Tutorials & Videos

🎓 Course Playlists

• AWS Strands Course: Complete 8-lesson course on building AI agents with AWS Strands SDK (watch the playlist)
• Voice Agents Playlist: Tutorials on building voice agents

🔧 Framework Tutorials

• AI Agents, MCP and more...: Mixed tutorials and project demos
• Build AI Agents: General AI agent development tutorials
• Build with MCP: Model Context Protocol tutorials and examples

⚡ Quick Start

You need Python 3.10+, uv or pip, and the API keys listed in the project's README. Most projects run on Nebius Token Factory.

# 1. Clone the repo and pick a project
git clone https://github.com/Arindam200/awesome-ai-apps.git
cd awesome-ai-apps/starter_ai_agents/agno_starter

# 2. Add your API keys
cp .env.example .env

# 3. Install dependencies
uv sync                              # or: pip install -r requirements.txt

# 4. Run it (each project's README has the exact command)
uv run main.py                       # or: python main.py / streamlit run app.py

🤝 Contributing

We welcome contributions from the community! Here's how you can help:

• 💡 Add a new project: a self-contained, runnable example with its own README, dependencies and .env.example
• 🔧 Fix issues: code improvements and bug fixes for existing projects
• 📝 Improve documentation: help make projects easier to set up and understand
• 🐛 Report bugs or suggest improvements via GitHub Issues

Before you open a PR, read the Contributing Guidelines, check existing issues to avoid duplicates, and follow the project README template. This project follows a Contributor Code of Conduct; by participating, you agree to abide by its terms.

🌟 Contributors

Awesome AI Apps is built by its community. Thank you to everyone who has shipped a project, fixed a bug or improved the docs! 💜

Want to see your avatar here? Pick an open issue or add a project.

👥 Maintainers

For questions, suggestions or partnerships, reach out to the maintainers on GitHub or at contact@studio1hq.com.

📬 Stay Updated

Easy-to-follow weekly tutorials and deep dives on AI, LLMs and agent frameworks, for developers who want to learn, build and stay ahead. Subscribe to our newsletter!

📜 License

This repository is licensed under the MIT License. Feel free to use and modify the examples for your projects.

⭐ Star History

If this collection helped you, consider giving it a ⭐. It helps other developers find it.