Learn Agentic AI

This repo is part of the Panaversity Certified Agentic & Robotic AI Engineer program. It covers AI-201 and AI-202 courses.

Watch The NVIDIA CEO Jensen Huang Keynote at CES 2025

Reference:

https://www.linkedin.com/posts/alexwang2911_aiagents-robotics-technology-activity-7282829390445453314-QLeS

AI-201: Fundamentals of Agentic AI - From Foundations to Autonomous Agents

AI 201 Fundamentals of Agentic AI we cover chapters: 00-06, 7b, 8, 8b, 9, 9b, 10, 10b, and 11

Kickstart your journey into Agentic AI! This course provides a rapid yet comprehensive introduction to Conversational, Generative, and Agentic AI. You'll master the foundational concepts, then immediately build practical Conversational AI applications to understand human-AI interaction firsthand. The focus quickly shifts to Agentic Design Patterns, which you'll implement using CrewAI to create truly autonomous AI agents. You'll become proficient with CrewAI, developing agents ready for real-world tasks. Furthermore, you'll gain the unique skills to construct Model Context Protocol (MCP) servers and agents, enabling you to build next-generation augmented LLMs. Finally, we'll explore the groundbreaking potential of Agentic Payments, envisioning the future of AI in finance.

AI-201 Video Playlist

Note: These videos are for additional learning, and do not cover all the material taught in the onsite classes.

AI-202: Advanced Agentic AI Engineering - Master Enterprise-Scale AI Agent Development

AI 202 Advanced Agentic AI we cover chapters: 06, 7a, 8, 8a, 9, 9a, 10, 10a, 11, and 12

Ready to engineer truly sophisticated AI agent systems? AI-202 builds upon your AI-201 foundation to propel you into advanced Agentic AI engineering. You'll master powerful frameworks like Microsoft AutoGen to construct complex agents for intricate tasks and advanced decision-making. Focusing on Agent-to-Agent communication and orchestration, you'll develop enterprise-ready multi-agent solutions. You'll build robust Model Context Protocol (MCP) servers, and then craft dynamic, user-centric agentic frontends with Next.js and TypeScript. The course culminates in a professional project where you'll design and deploy a complete enterprise-grade agentic solution, showcasing your mastery of cutting-edge AI technologies and your readiness for the forefront of the field.

Which is the Best Agentic Framework?

Let's break down CrewAI, Autogen, and LangGraph for agentic development to help you decide which might be "better" for your needs. There isn't a single "best" – it highly depends on what you prioritize in your agentic development project.

Here's a comparative look across several key dimensions:

1. Core Philosophy & Structure:

• CrewAI: Focuses on structured teams of agents with specialized roles. It's designed to mimic human teams, where each agent ("crew member") has a specific task and collaborates towards a common goal. Think of it as organizing agents into a well-defined project team.

  • Key Concept: Crews, Agents with Roles, Tasks, Collaboration within teams.
  • Analogy: Building a small company with specialized departments working together.

• Autogen: Emphasizes conversational agents and flexible, dynamic multi-agent conversations. It's about creating agents that can fluidly interact, debate, and refine ideas through natural language conversations. Think of it as setting up a dynamic forum or discussion group of agents.

  • Key Concept: Conversational Agents, Multi-Agent Chat, Dynamic Interaction, Flexibility.
  • Analogy: Setting up a lively debate or brainstorming session among experts.

• LangGraph: Prioritizes building robust, stateful, and reliable agentic workflows as graphs. It offers a more programmatic and state-centric approach, allowing you to define complex workflows and manage state transitions explicitly. Think of it as architecting a complex, reliable system with agents as nodes in a workflow graph.

  • Key Concept: Workflows as Graphs, Stateful Agents, Robustness, Reliability, Explicit State Management.
  • Analogy: Designing a complex assembly line or a robust data processing pipeline with agent components.

2. Ease of Use & Learning Curve:

• CrewAI: Relatively easy to get started with, especially if you understand the concept of teams and roles. The API is designed to be intuitive for defining agents, tasks, and crews. Good for beginners wanting to structure agent collaborations quickly.

  • Learning Curve: Gentle to Moderate. Focuses on team structure, which is conceptually straightforward.

• Autogen: Also relatively easy to begin with, particularly for conversational scenarios. Setting up basic agent chats is straightforward. Complexity increases as you dive deeper into more nuanced agent behavior and conversation management.

  • Learning Curve: Gentle to Moderate. Easy start, but mastering complex conversations requires more effort.

• LangGraph: Steeper learning curve initially. The graph-based workflow approach is more programmatic and requires understanding state management and graph concepts. Offers more control but demands a more engineering-oriented mindset.

  • Learning Curve: Moderate to Steep. Requires understanding graph workflows and state management.

3. Flexibility & Customization:

• CrewAI: Structured but offers good flexibility within its team-based paradigm. You can customize agent roles, tasks, tools, and collaboration strategies within a crew. Less flexible if you want to deviate significantly from the team structure.

  • Flexibility: Medium. Excellent within its team-based structure, less so outside of it.

• Autogen: Highly flexible and customizable for conversational flows. You have a lot of control over agent behavior, prompts, conversation styles, and integration of tools. Adaptable to a wide range of conversational and interactive agent scenarios.

  • Flexibility: High. Very adaptable for diverse conversational and interactive agent needs.

• LangGraph: Extremely flexible and highly customizable for workflow design. You have granular control over every node (agentic step) in the graph, state transitions, error handling, and more. Best for building very specific and complex agentic workflows with precise control.

  • Flexibility: Very High. Offers the most granular control over complex workflow design.

4. Complexity Handling:

• CrewAI: Good for managing moderate complexity, especially in task delegation and team coordination. The crew structure helps break down complex projects into manageable agent tasks within a team. Can become less manageable for very intricate workflows beyond team interactions.

  • Complexity Handling: Moderate. Effective for team-based project breakdown, less so for highly intricate workflows.

• Autogen: Handles conversational complexity well, especially in dynamic multi-agent discussions. Can manage complex interactions and idea refinement through conversation. May become challenging to track state and manage long, branching conversations without explicit state management.

  • Complexity Handling: Moderate to High for conversational complexity. Can become complex to manage state in long conversations.

• LangGraph: Designed explicitly for handling high complexity in workflows and state management. The graph structure allows you to model and manage intricate, multi-step agentic processes with clear state transitions, error handling, and branching logic. Best for truly complex, production-ready agent systems.

  • Complexity Handling: High to Very High. Excellent for managing intricate workflows, state, and complex system architectures.

5. Collaboration & Teamwork (Agentic):

• CrewAI: Strongest focus on agent collaboration and teamwork. It's built around the concept of crews working together, making it ideal for scenarios where agents need to collaborate closely to achieve a shared objective.

  • Collaboration Focus: Very High. Core strength is team-based agent collaboration.

• Autogen: Supports multi-agent conversations and collaboration, but more in a dynamic, less structured way. Agents can debate, discuss, and refine ideas, but the collaboration is driven by conversation rather than a predefined team structure.

  • Collaboration Focus: Medium to High. Supports conversation-driven collaboration in a flexible manner.

• LangGraph: Supports collaboration implicitly through workflow design. You can design workflows where agents interact and exchange information as part of a larger process. Collaboration is defined programmatically within the graph, rather than being the central organizing principle like in CrewAI.

  • Collaboration Focus: Medium. Collaboration is supported as part of workflow design, not the primary focus.

6. Scalability & Robustness:

• CrewAI: Scalability and robustness depend on how well you design your crews and agents. Can scale if you effectively parallelize tasks and manage resource usage within your crews. Robustness needs to be considered in agent design and task handling.

  • Scalability/Robustness: Medium. Requires design considerations for scalability and robustness.

• Autogen: Scalability depends on conversation management and resource usage. Handling very large numbers of concurrent conversations or very long conversations might require careful management. Robustness depends on agent behavior and error handling in conversations.

  • Scalability/Robustness: Medium. Scalability and robustness require management of conversational complexity.

• LangGraph: Designed with scalability and robustness in mind. The graph-based approach allows for building more robust and maintainable agentic workflows. Explicit state management, error handling, and observability features contribute to building reliable systems.

  • Scalability/Robustness: High. Designed for building robust and scalable agentic systems.

7. Specific Use Cases:

• CrewAI:

  • Project-based agentic tasks: Research projects, content creation, structured problem-solving where tasks can be delegated to specialized agents within a team.
  • Simulating human teams: Scenarios where you want to mimic the workflow of a human project team with specialized roles.
  • Learning about structured agent collaboration: Good starting point for understanding agentic teamwork.

• Autogen:

  • Conversational agents and assistants: Building chatbots, interactive agents, agents for debate, brainstorming, and idea generation.
  • Multi-agent simulations and games: Scenarios where you want to simulate dynamic interactions between agents.
  • Exploring emergent behavior in agent conversations: Researching how agents interact and evolve ideas through dialogue.

• LangGraph:

  • Complex, production-ready agentic systems: Building robust applications requiring intricate workflows, state management, and reliability (e.g., complex automation, data processing pipelines).
  • Systems with well-defined workflows: Applications where the agentic process can be clearly modeled as a workflow graph with distinct steps and state transitions.
  • Applications needing strong state management and error handling: Scenarios requiring precise control over state and robust error recovery in agentic workflows.

Summary Table:

Feature	CrewAI	Autogen	LangGraph
Core Philosophy	Structured Teams	Conversational Agents	Workflow Graphs
Ease of Use	Moderate	Moderate	Moderate to Steep
Flexibility	Medium (Team-centric)	High (Conversation-centric)	Very High (Workflow-centric)
Complexity Handling	Moderate (Team-based)	Moderate to High (Conversation-based)	High to Very High (Workflow & State-based)
Collaboration	Very High (Teamwork)	Medium to High (Conversational)	Medium (Workflow-integrated)
Scalability/Robustness	Medium (Design-dependent)	Medium (Conversation Management)	High (Designed for Robustness)
Best For	Team-based Projects, Structured Tasks	Conversational Agents, Dynamic Interaction	Complex Workflows, Production Systems

Which is "Better"?

There's no universally "better" choice. It boils down to your specific project goals and priorities:

• Choose CrewAI if: You want to build agents that work together as structured teams to solve problems, and you value clear roles and team-based collaboration. It's a good entry point to structured agentic development.

• Choose Autogen if: You are focused on creating conversational agents that can interact dynamically, debate, and refine ideas through natural language. It excels in building interactive and conversational AI.

• Choose LangGraph if: You are building complex, production-grade agentic systems that require robust workflows, explicit state management, and high reliability. It's the best choice for intricate, scalable, and dependable agentic applications.

Recommendation:

• Start with CrewAI or Autogen if you are new to agentic development and want a quicker ramp-up. Experiment with their more intuitive APIs.
• Consider LangGraph if you anticipate building complex, production-ready systems from the outset, or if you find the need for robust state management and workflow control becomes critical as your projects evolve.

Ultimately, the best way to decide is to experiment with each framework on a small project relevant to your needs. Try building a simple application with each to get a feel for their strengths and weaknesses in practice. You might even find that a combination of these frameworks could be beneficial for certain complex projects.

Which Agentic Framework is Being Used the Most?

Determining which of these frameworks is "used the most" is tricky and depends on how we define "used." There isn't a single, definitive public dashboard that tracks usage across all these tools perfectly. However, we can use several indicators to get a reasonable picture.

Here's an analysis based on publicly available data and general observations in the AI agent development community:

Indicators of Usage & Popularity (and what they suggest):

1. GitHub Stars and Forks:

   • Generally reflects community interest and adoption. Higher stars often correlate with broader awareness and usage, though it's not a perfect measure of active deployments.

   • Likely Ranking (based on current GitHub numbers, which fluctuate):

     • Autogen: Likely to be highest. Autogen, being backed by Microsoft and focusing on flexible conversation-based agents, has garnered significant attention and a large number of GitHub stars and forks.
     • CrewAI: Likely to be second highest and rapidly growing. CrewAI, while newer, has a very clear and appealing concept (team-based agents), which resonates strongly, leading to rapid growth in GitHub stars.
     • LangGraph: Likely to be lower than Autogen and CrewAI, but still significant. LangGraph is more specialized and targets robust, production-ready systems, which might appeal to a slightly more niche but serious developer audience.

2. PyPI Download Statistics:

   • Shows how often the libraries are being installed. Higher downloads can suggest broader usage, but also include installations for experimentation, tutorials, and CI/CD.

   • Likely Ranking (Less publicly available real-time data, but general trends suggest):

     • Autogen: Likely to be highest download count. Broader appeal and more introductory tutorials might lead to more initial installations.
     • CrewAI: Likely to be second highest and growing fast. Increasing popularity means more installations for projects and experimentation.
     • LangGraph: Likely to be lower, potentially reflecting its more specialized use case and potentially later entry to the scene compared to Autogen.

3. Community Activity (Discord, Forums, Social Media):

   • Indicates active user base, support network, and general buzz. Larger and more active communities often suggest wider adoption and support.

   • Likely Ranking (Qualitative observation):

     • Autogen: Likely to have a very large and active community. Being from Microsoft and targeting a broad audience, it likely benefits from a larger initial community and resources.
     • CrewAI: Likely to have a rapidly growing and very engaged community. The focused nature of CrewAI attracts developers interested in structured agentic teams, creating a strong community.
     • LangGraph: Likely to have a smaller but potentially very focused and technically deep community. Its more specialized nature might lead to a community of developers tackling complex, production-oriented problems.

4. Mentions in Tutorials, Blog Posts, and Courses:

   • Reflects educational resources and ease of learning, which can drive adoption. More readily available learning materials can lower the barrier to entry.

   • Likely Ranking (Observation based on online content search):

     • Autogen: Likely to have the most tutorials and online content. Its broader appeal and conversational focus make it a popular topic for demos and tutorials.
     • CrewAI: Likely to have a rapidly increasing number of tutorials. Its straightforward concept makes it easy to demonstrate and teach.
     • LangGraph: Likely to have fewer introductory tutorials compared to Autogen and CrewAI but more in-depth, technically focused documentation and examples aimed at production use cases.

5. Anecdotal Evidence & General Perception in the AI Community:

   • Informal understanding based on conversations, online discussions, and industry trends.

   • Likely Perception:

     • Autogen: Widely recognized and talked about, often seen as a versatile framework for various agentic tasks, especially conversational.
     • CrewAI: Gaining significant momentum and being recognized as a powerful and intuitive framework for structured agentic collaboration and team-based tasks.
     • LangGraph: Perceived as the go-to choice for more complex, robust, and production-ready agent systems, though perhaps less widely adopted initially due to its steeper learning curve and specialized focus.

Overall Conclusion based on Indicators:

Based on the available indicators, it's highly likely that Autogen is currently the most "used" in terms of sheer number of projects and broader initial adoption. This is likely due to:

• Microsoft backing and visibility.
• Focus on conversational agents, a very popular and accessible area in AI.
• Versatility and flexibility for a wide range of agentic tasks.
• Strong initial community and resources.

However, CrewAI is rapidly gaining popularity and momentum. It's likely to be the fastest-growing framework in terms of adoption, especially for projects focused on structured agentic collaboration and team-based workflows. Its clear and intuitive concept makes it very appealing.

LangGraph, while likely having a smaller overall user base currently, is strategically positioned for serious, production-oriented agent development. Its focus on robustness, state management, and complex workflows makes it essential for building reliable and scalable agentic systems, even if its initial adoption numbers might be lower compared to Autogen and CrewAI.

Important Nuances:

• "Most Used" can change quickly. The AI agent landscape is rapidly evolving. Popularity can shift as frameworks mature and new needs emerge.
• Different Use Cases: Each framework excels in different areas. "Most used overall" doesn't mean "best for every use case." LangGraph might be "most used" for production systems needing robustness, even if Autogen is "most used" in general experimentation and tutorials.
• Maturity: Autogen has been around longer in the open-source space, giving it a head start. CrewAI and LangGraph are newer but quickly catching up.

In summary:

If we had to rank them by current likely broadest usage, it would probably be:

1. Autogen (Likely broadest current adoption)
2. CrewAI (Rapidly growing, catching up fast, very strong momentum)
3. LangGraph (More specialized, solid adoption for production systems, but potentially smaller overall user base currently)

It's crucial to choose the framework that best aligns with your specific project requirements rather than just chasing "the most popular" one. Each of these frameworks offers unique strengths for different types of agentic development.