AI Agents Atlas: 500+ AI Agent Projects Directory

The repository includes a wide variety of LangGraph-based agent implementations covering several specialized domains. You can use these as templates or reference implementations for:

Workflow Orchestration

• Multi-Agent Workflow (Supervisor): A supervisor agent orchestrating multiple specialized agents.
• Hierarchical Agent Teams: A top-level supervisor that delegates tasks to specialized sub-agents.
• Multi-Agent Collaboration: Multiple specialized agents working together on complex tasks.
• Plan-and-Execute Agent: An agent that generates a multi-step plan and then executes it sequentially.
• Reflection/Reflexion Agents: Agents that critique, revise, or reflect on their own outputs/actions for iterative improvement.

Information Retrieval (RAG)

• Adaptive RAG: Dynamic retrieval that adjusts based on query complexity.
• Agentic RAG: An agent that determines the best retrieval strategy before generating a response.
• Corrective RAG (CRAG): Evaluates and refines retrieved documents before generation.
• Self-RAG: A system that reflects on responses and retrieves additional information if needed.
• Local RAG: Adaptive and Self-RAG implementations designed for use with local models.

Specialized Agents

• SQL Agent: An agent designed to answer questions about SQL databases.
• Code Assistant: A resilient assistant with error checking and iterative refinement.
• Customer Support Agent: A graph-based agent for handling customer inquiries.
• Data Extraction: Robust extraction using retry mechanisms.

The repository provides several specialized agent implementations using the Microsoft AutoGen framework. These are available as Jupyter Notebooks for various industries and technical tasks:

Information Retrieval & Tools

• Web Search: Solves tasks requiring real-time web information.
• Tool Integration: Demonstrates using pre-provided tools as callable functions.
• RAG Group Chat: Enables collaborative group chats using Retrieval Augmented Generation.

Multimodal & Audio

• Audio Processing: Uses Whisper for transcription and translation.
• Multimodal Agents: Implementations using DALLE and GPT-4V, Llava, or GPT-4V for visual and conversational interactions.

Database & Management

• SQL: Converts natural language into SQL queries.
• AgentEval: A multi-agent assessment system for evaluating LLM application utility.
• AgentOps: Monitoring and analytics for tracking LLM calls, tool usage, and errors.
• AgentBuilder: Automatically builds multi-agent systems.

The repository provides several AutoGen implementation patterns via Jupyter Notebooks, covering:

Code Generation & Retrieval

• Automated Task Solving: Code generation, execution, and debugging.
• RAG (Retrieval Augmented Generation): Using retrieval-augmented agents for Q&A, including Qdrant-based retrieval.

Multi-Agent Collaboration

• Group Chat: Managing group tasks with a manager, data visualization via group chat, and complex task solving with large groups (e.g., 6 members).
• Advanced Patterns: Using Graph Transition Paths (Finite State Machines), custom speaker selection functions, and SocietyOfMindAgent for simulating inner-monologues.

Workflow Patterns

• Sequential Chats: Single-initiator sequential tasks, asynchronous sequential tasks, and sequential chats with different initiating agents.
• Nested Chats: Hierarchical problem solving using nested chats, including sequential nested chats and specific industry applications like supply chain optimization or conversational chess.

Agno is a lightweight, fast agent framework suitable for single-agent tools and rapid prototyping. The following Python-based agent implementations are available:

Finance & Research

• Finance Agent (Thinking): Real-time stock insights and analyst recommendations.
• Financial Reasoning Agent: Claude 3.5 Sonnet-based analysis using Yahoo Finance data.
• Research Scholar Agent: Advanced academic searches and publication analysis.
• DeepKnowledge: Iterative search through knowledge bases with deep reasoning.

Media & Content

• YouTube Agent: Summarizes videos, identifies timestamps, and themes.
• Media Trend Analysis Agent: Analyzes emerging trends and influencers.
• Movie Recommendation Agent: Personalized suggestions using Exa and GPT-4o.
• Book Recommendation Agent: Personalized suggestions using literary data.

Specialized Assistants

• Support Agent: Real-time answers and code examples for the Agno framework.
• Study Partner: Creates study plans and finds educational resources.
• Shopping Partner Agent: Product recommendations from Amazon and Flipkart.
• Recipe Creator: Personalized recipes based on ingredients.
• Readme Generator Agent: Generates GitHub READMEs.
• Legal Document Analysis Agent: Analyzes legal PDFs using vector embeddings.
• MCP Airbnb Agent: Searches Airbnb listings using MCP and Llama 4.

The Meeting Notes Agent converts meeting transcripts into structured markdown notes containing an executive summary, key decisions, action items (with owners and due dates), blockers, and the next meeting time.

You can run the agent using several input methods:

• Built-in sample: Run without arguments to use the included sample transcript.
• Transcript file: Provide a path to a .txt file using the --transcript flag.
• Inline text: Provide raw transcript text directly using the --text flag.

You can also specify a custom destination for the output using the --output flag.

# Use built-in sample transcript
python agent.py

# Your own transcript file
python agent.py --transcript meeting_transcript.txt

# Inline text
python agent.py --text "Alice: Let's ship by Friday. Bob: I need 2 more days for testing..."

# Custom output path
python agent.py --transcript meeting_transcript.txt --output sprint_notes.md