LLM Agents & Ecosystem Handbook
A unified handbook for building, deploying and understanding LLM agents and the wider ecosystem
A polished, curated collection of Large Language Model (LLM) agents, tutorials and ecosystem insights. This handbook highlights projects that push the boundaries of generative AI, multi-agent collaboration, retrieval-augmented generation (RAG), voice and game agents, and more. It goes beyond simple link aggregation, aiming to be a one-stop reference for building, deploying, and understanding LLM applications across the entire stack.
Tip: If you enjoy this list, please consider starring the repository to help others discover it!
Table of Contents
- Top Agent Frameworks
- Agent Toolkits & Platforms
- Starter AI Agents
- Advanced AI & Domain-Specific Agents
- Multi-Agent Teams
- Voice & Game Agents
- RAG & Memory Examples
- MCP Agent Integrations
- LLM Evaluation Frameworks
- Example Projects
- Tutorials & Learning Resources
- Other Educational Spaces
- Unique Features
- Languages & Multilingual Support
- Interactive Demos & Resources
- Datasets & Design Assets
- Documentation & Roadmap
- Complete Applications
- Beginner’s Guide
- Contributing
- License
- Maintainer
Why this repository stands out
This curated collection aims to be a comprehensive resource for developers and researchers building their own LLM applications. In addition to code examples, it provides:
- Comparative analysis of leading agent frameworks: A quick matrix contrasting frameworks (LangGraph, AutoGen, CrewAI, Smolagents, etc.) with key features to help you choose.
- Guidance on framework selection: Practical advice based on task complexity, collaboration needs and ecosystem integrations.
- LLM evaluation toolbox: Summaries of tools like Promptfoo, DeepEval, MLflow LLM Evaluate, RAGAs and Langfuse to measure performance and safety.
- 60+ skeleton projects: The
agentsfolder contains scaffolded agents across many domains (blogging, medical imaging, music generation, multimodal input, news, finance, research, scraping, consultancy, system design, compliance, marketing, scheduling, supply-chain, healthcare, education). Each skeleton includes aREADME.mdandmain.py. - Agent skeleton generator:
scripts/create_agent.pyto spin up new agent skeletons in seconds.
Top Agent Frameworks
| Framework | Description & Key Features |
|---|---|
| LangGraph | Graph/DAG-based orchestration for complex multi-step workflows. |
| OpenAI Agents SDK | Structured runtime with tool-calling and role-based agents. |
| AutoGen (AG2) | Event-driven multi-agent conversations and human-in-the-loop. |
| CrewAI | Role-based “crew” collaboration with memory and error handling. |
| Google AgentKit (ADK) | Modular Gemini/Vertex AI agent kit with hierarchical tools. |
| Dify | Low-code builder with RAG and function calling. |
| LangChain & Tools | Mature chains, memory and 3rd-party integrations. |
| Smolagents | Minimal, code-centric loop (agents write & execute code). |
| Semantic Kernel | .NET-first skills/plans; enterprise-friendly. |
| LlamaIndex Agents | Retrieval-focused agents for data-heavy apps. |
| Strands Agents | Provider-agnostic SDK with OpenTelemetry. |
| Pydantic AI | Type-safe IO/tool signatures with great DX. |
Agent Toolkits & Platforms
| Project | Description |
|---|---|
| AutoGPT | Toolkit for autonomous agents (creation, benchmarking, UI/CLI). |
| Ollama | Run LLMs locally across macOS/Windows/Linux/Docker. |
| Lobe Chat | Open-source chat UI with plugins and multimodal support. |
| OpenDevin | Open initiative towards an AI software engineer. |
| Open Interpreter | Natural-language coding & local computer control. |
| MetaGPT | Multi-agent “virtual company” for complex tasks. |
| PrivateGPT | Secure offline Q&A over your documents. |
| GPT-Engineer | From natural-language spec to code. |
| LlamaIndex Tools | Connectors/tools for data agents. |
| Flowise | Drag-and-drop builder for LLM workflows. |
| FastChat | Train/serve/evaluate chatbots. |
| Mem0 | Memory layer for personalised LLMs. |
| Cal.ai | Scheduling assistant with email handling. |
| Aider | CLI pair-programming agent with Git integration. |
| Jan | Offline ChatGPT-style desktop app. |
Starter AI Agents
| Agent | Description |
|---|---|
| AI Blog to Podcast Agent | Convert blog posts into podcasts. |
| AI Data Analysis Agent | Insights from CSV/structured data. |
| AI Travel Agent | Trip itineraries (local/cloud). |
| AI Music Generator | Compose via generative models. |
| AI Meme Generator (Browser) | Creates memes by overlaying captions on images. |
| AI Breakup Recovery Agent | Supportive advice for emotional situations. |
| AI Health & Fitness Agent | Health metrics & coaching. |
| Gemini Multimodal Agent | Text+image multimodal demo. |
Advanced AI & Domain-Specific Agents
| Agent | Description |
|---|---|
| AI Deep Research Agent | Multi-source research & synthesis. |
| AI Consultant Agent | Domain-expert strategy & advice. |
| AI System Architect Agent | From requirements to architecture. |
| AI Lead Generation Agent | Identify & qualify prospects. |
| AI Meeting Agent | Summaries & action items. |
| OpenAI Research Agent | Research workflows with tools. |
| Explainable AI Finance Agent | Finance with interpretability. |
| Web Scraping Agent | Crawl & extract structured data. |
| Document Processing Agent | OCR + analysis/summarisation. |
| Sentiment Analysis Agent | Classify sentiment at scale. |
| Technical Translation Agent | Preserve domain terminology. |
| Research Synthesizer Agent | RAG + coherent reporting. |
Multi-Agent Teams
| Team | Description |
|---|---|
| Competitor Intelligence Team | Market/competitor research & reporting. |
| Finance Agent Team | Budgeting, forecasting and reporting. |
| Teaching Agent Team | Lesson planning, delivery and assessment. |
| Multi-Agent Team Demo | Cross-role collaboration patterns. |
| Mixture of Agents Demo | Specialised agent ensembles. |
Voice & Game Agents
| Agent | Description |
|---|---|
| Voice Summary Agent | Transcribe and summarise audio. |
| AI Audio Tour Agent | Generate audio tours for museums/cities. |
| Customer Support Voice Agent | Handle spoken queries and log issues. |
| Voice RAG Agent | Voice input + retrieval + TTS. |
| Tic-Tac-Toe Agent | Autonomous gameplay template. |
RAG & Memory Examples
| Example | Description |
|---|---|
| Agentic RAG with Reasoning | Retrieval → reasoning → generation. |
| Hybrid Search RAG | Vector + keyword retrieval. |
| Vision RAG | Apply RAG to visual data. |
| CRAG (Corrective RAG) | Human-feedback corrective loop. |
| Local RAG Agent | Fully offline retrieval pipeline. |
MCP Agent Integrations
| Agent | Description |
|---|---|
| Browser MCP Agent | Drive a browser (search/click/forms). |
| GitHub MCP Agent | Read/write/manage repositories. |
| Notion MCP Agent | Create/update/query Notion pages & DBs. |
| Travel Planner MCP Agent Team | Multi-agent trip planning via MCP. |
LLM Evaluation Frameworks
See evaluation_frameworks/README.md for details about Promptfoo, DeepEval, MLflow LLM Evaluate, RAGAs, Deepchecks, LangSmith, TruLens, Arize Phoenix and Langfuse.
Example Projects
The agents directory contains many agent skeletons organised by category. Category folders (starter, advanced, teams, rag) act as indexes, while each agent lives in its own top-level folder under agents/.
Quick picks
- Summarization Agent – agents/summarization_agent
- Data Analysis Agent – agents/data_analysis_agent
- Travel Itinerary Agent – agents/travel_itinerary_agent
- Voice Assistant Demo – agents/voice_agent_demo
- Meme Generator Agent – agents/meme_generator_agent
- Health & Fitness Agent – agents/health_fitness_agent
- Breakup Recovery Agent – agents/breakup_recovery_agent
- AI Blog to Podcast Agent – agents/ai_blog_to_podcast_agent
- AI Medical Imaging Agent – agents/ai_medical_imaging_agent
- AI Music Generator Agent – agents/ai_music_generator_agent
- Local News Agent – agents/local_news_agent
- Gemini Multimodal Demo – agents/gemini_multimodal_agent_demo
Tutorials & Learning Resources
Hands-on tutorials live in tutorials/ directory.
- RAG Tutorials →
tutorials/rag_tutorials - Memory Apps Tutorials →
tutorials/memory_apps - Chat with X Tutorials →
tutorials/chat_with_x_tutorials - Fine-Tuning Tutorials →
tutorials/fine_tuning_tutorials
Other Educational Spaces
- Interactive demos & notebooks:
web_apps,notebooks - Datasets & design assets:
datasets,design - LLM ecosystem overview:
ecosystem/overview.md - Complete applications:
complete_apps
Unique Features
- Educational focus: Detailed tutorials (RAG, memory, chat with X, fine-tuning) + 60+ scaffolded agents.
- Framework comparison & guidance: Practical, vendor-neutral advice.
- Agent skeleton generator:
scripts/create_agent.py. - Evaluation toolbox: Promptfoo, DeepEval, RAGAs, etc.
- Ecosystem breadth: Training, tools, production, local inference, operations, interpretability.
- Community roadmap: See
docs/roadmap.md.
Languages & Multilingual Support
We welcome translations. See TRANSLATION.md. A Technical Translation Agent exists under agents/technical_translation_agent.
Interactive Demos & Resources
Web Apps
- Streamlit Summariser –
web_apps/streamlit_summarizer - Gradio FAQ Bot –
web_apps/gradio_faq_bot
Jupyter Notebooks
- Getting Started –
notebooks/getting_started.ipynb
Datasets & Design Assets
- Sample dataset –
datasets/sample_products.csv(+datasets/README.md) - Architecture diagram –
design/architecture_diagram.png(+design/README.md)
Documentation & Roadmap
- Best practices –
docs/best_practices.md - Framework comparison –
docs/framework_comparison.md - Evaluation frameworks guide –
evaluation_frameworks/README.md - Quickstart –
tutorials/quickstart.md - Roadmap –
docs/roadmap.md - Changelog –
CHANGELOG.md
Complete Applications
- Task Planner –
complete_apps/task_planner - Health Coach –
complete_apps/health_coach
Beginner’s Guide
If you’re new to LLMs, start here → docs/beginners_guide.md
Contributing
Contributions are welcome! Please open an issue or PR. Ensure new entries are permissively licensed (MIT/Apache-2.0) and well-documented.
License
MIT — see LICENSE.
Maintainer
Curated & maintained by Sayed Allam (oxbshw). If you find this helpful, please ⭐ star the repo and share feedback via issues/PRs.