Agentic AI Learning Roadmap (with GitHub Repos)

Here’s a clear, practical Agentic AI Learning Roadmap you can follow from beginner → advanced, inspired by modern agentic systems (planning, tool use, memory, autonomy), but structured so you can actually build things step by step.

Phase 1: Foundations of Agentic AI

Goal: Understand agents vs chatbots and build your first goal-driven system.

Learn

• GenAI vs Agentic AI
• Prompting for reasoning
• LLM APIs & function calling

Hands-On GitHub Repos

• OpenAI Cookbook https://github.com/openai/openai-cookbook → Function calling, tool use, reasoning patterns
• LangChain Basics https://github.com/langchain-ai/langchain → Chains, tools, memory fundamentals

Build

Goal-based task solver (input goal → output steps → execute)

Phase 2: Reasoning & Planning Agents

Goal: Teach agents to think, plan, and self-correct.

Learn

• ReAct
• Chain-of-Thought
• Planner–Executor
• Self-reflection loops

Hands-On GitHub Repos

• ReAct Prompting https://github.com/ysymyth/ReAct → Core ReAct pattern
• Tree of Thoughts https://github.com/princeton-nlp/tree-of-thought-llm → Advanced reasoning & search
• LangGraph https://github.com/langchain-ai/langgraph → Agent state machines & planning

Build

AI trip planner Bug-fixing agent

Phase 3: Tool-Using Agents

Goal: Enable agents to take real actions.

Learn

• Tool schemas
• API calls
• Code execution
• Error recovery

Hands-On GitHub Repos

• OpenAI Function Calling Examples https://github.com/openai/openai-cookbook/tree/main/examples → Tool orchestration
• LangChain Tools https://github.com/langchain-ai/langchain/tree/master/libs/langchain/langchain/tools
• Browser Use Agent https://github.com/browser-use/browser-use → Web-browsing agents

Build

AI research agent Web automation agent

Phase 4: Memory & Knowledge (RAG)

Goal: Persistent, learning agents.

Learn

• Vector DBs
• Long-term memory
• RAG pipelines
• Knowledge graphs

Hands-On GitHub Repos

• LlamaIndex https://github.com/run-llama/llama_index → Best RAG + agent memory framework
• RAG Examples https://github.com/langchain-ai/langchain/tree/master/templates
• Weaviate RAG https://github.com/weaviate/weaviate-examples

Build

Personal memory agent Document-aware assistant

Phase 5: Agent Frameworks

Goal: Production-grade agent systems.

Learn

• Agent orchestration
• State & workflow control
• Scaling agents

Hands-On GitHub Repos

• CrewAI https://github.com/joaomdmoura/crewAI → Role-based multi-agent systems
• Microsoft AutoGen https://github.com/microsoft/autogen → Conversational & tool-using agents
• LangGraph Agents https://github.com/langchain-ai/langgraph/tree/main/examples

Build

AI customer support agent Autonomous workflow agent

Phase 6: Multi-Agent Systems (MAS)

Goal: Agents that collaborate, debate, and supervise.

Learn

• Collaborative agents
• Supervisor agents
• Debate & consensus
• Swarm patterns

Hands-On GitHub Repos

• AutoGen Multi-Agent Demos https://github.com/microsoft/autogen/tree/main/samples
• CrewAI Multi-Agent Examples https://github.com/joaomdmoura/crewAI-examples
• MetaGPT https://github.com/geekan/MetaGPT → Software-engineering agent teams

Build

Multi-agent research team AI software dev team

Phase 7: Evaluation, Safety & Guardrails

Goal: Make agents safe, reliable, and measurable.

Learn

• Agent evaluation
• Guardrails
• Observability
• Cost & latency control

Hands-On GitHub Repos

• LangSmith https://github.com/langchain-ai/langsmith-sdk
• TruLens https://github.com/truera/trulens
• RAGAS https://github.com/explodinggradients/ragas
• Arize Phoenix https://github.com/Arize-ai/phoenix

Build

Evaluated production agent Agent monitoring dashboard

Phase 8: Advanced & Real-World Agents

Goal: Autonomous systems at scale.

Learn

• Self-improving agents
• Feedback loops
• Workflow engines
• Agents + SaaS

Hands-On GitHub Repos

• OpenHands (formerly OpenDevin) https://github.com/All-Hands-AI/OpenHands → Autonomous software engineer
• SuperAGI https://github.com/TransformerOptimus/SuperAGI → Autonomous agent platform
• AutoGPT (Reference) https://github.com/Significant-Gravitas/AutoGPT → Understand pitfalls of autonomy

Build

AI operations agent Autonomous startup agent