Explore AI Projects by Category
From single-file starter agents and conversational tool loops to multi-agent supervisor swarms, hybrid dense-sparse RAG, FastMCP servers, and low-latency voice pipelines.
Starter AI Agents
Beginner-friendly, single-file agents demonstrating fundamental tool-calling and prompting patterns.
Advanced AI Agents
Multi-agent systems, autonomous teams, hierarchical swarms, and complex workflow orchestrations.
RAG Tutorials & Pipelines
Retrieval Augmented Generation architectures, hybrid vector-BM25 search, reranking, and contextual retrieval.
MCP AI Agents
Implementations of the Model Context Protocol (MCP) connecting agents to external servers and local tools.
Agent Skills & Evaluations
Modular agent capabilities, structured reasoning protocols, reflection loops, and evaluation suites.
Chat With X Tutorials
Conversational interfaces grounded in databases, PDFs, documents, media streams, and external APIs.
Voice AI Agents
Low-latency, real-time speech-to-speech agents with streaming audio, VAD, and bidirectional WebSockets.
Generative UI & Frontends
Agents that dynamically render interactive frontend components, artifacts, and personalized user interfaces.
Always-On & Proactive Agents
Continuous background workers, proactive monitoring loops, scheduled automations, and event-driven triggers.
LLM Apps with Memory
Stateful architectures leveraging short-term working memory, semantic episodic memory, and long-term vector stores.
LLM Optimization & Evaluation
Tools for prompt optimization, structured schema enforcement, response caching, and evaluation metrics.
Framework Crash Courses
Structured tutorials exploring OpenAI Agents SDK, Google ADK, Smolagents, LangGraph, and CrewAI.
Game Playing Agents
LLM and reinforcement learning agents playing interactive games, chess, and simulation environments.
LLM Fine-Tuning
Guides and pipelines for fine-tuning open-source LLMs using LoRA, QLoRA, PEFT, and custom datasets.
How UtilityHub Categorizes AI Architecture
Rather than classifying projects solely by their marketing labels, UtilityHub evaluates the core execution model: whether state is cyclic (LangGraph/ReAct), conversational (AutoGen), protocol-isolated (FastMCP), or streaming (Voice/WebSockets). This helps developers find architectures that match their exact engineering constraints.