Coming soon
Agentic AI: Build AI Agents with Code
Build AI agents that use tools, remember context and finish multi-step tasks, with LangGraph and MCP.
- 2 months
- Intermediate to advanced (Python needed)
- 5 parts · 40 topics
Send us one message on WhatsApp, and we will tell you the launch date.
Planned syllabus
The full roadmap, basic to advanced
Every part builds on the one before it. You finish each part with a real project.
- Parts
- 5
- Duration
- 2 months
- Topics
- 40
- Projects
- 5
Part 1What an AI agent is
- Chatbot vs workflow vs agent
- The agent loop: think, act, observe
- The ReAct pattern
- When you should not use an agent
- Tool calling with the Claude and OpenAI APIs
- Writing tool descriptions the model understands
- Building an agent loop from scratch in Python
- Stop conditions and step limits
Project An agent that searches the web and writes a short report.
Part 2Tools and memory
- Tools: web search, calculator, file reader
- Calling real APIs as tools
- Running code safely in a sandbox
- Short-term memory: conversation state
- Long-term memory with a vector database
- Planning and breaking down tasks
- Handling tool errors and retries
- Structured state with Pydantic
Project A personal assistant that manages your to-do list in a database.
Part 3LangGraph and MCP
- LangGraph basics: nodes, edges, state
- Conditional edges and loops
- Checkpoints and resuming a run
- Human in the loop: approve before acting
- Streaming agent steps to the user
- Model Context Protocol (MCP): what it solves
- Using ready MCP servers
- Building your own MCP server
Project An agent that reads GitHub issues and drafts fixes for review.
Part 4Multi-agent systems
- Supervisor and worker agents
- Handoffs between agents
- CrewAI and the OpenAI Agents SDK: a quick look
- Claude Agent SDK basics
- Research agent: search, read, summarise
- Coding agent: plan, write, test
- Sharing state between agents
- Cost and speed trade-offs
Project A research team of 3 agents that writes a checked report.
Part 5Production
- Evaluating agents with test tasks
- Tracing every step with LangSmith
- Guardrails and permission limits
- Prompt injection through tools and web pages
- Budgets: token and step limits
- Deploying an agent behind an API
- Background jobs and queues
- Monitoring and alerts
Project Deploy your agent with a dashboard of every run.
This is the planned syllabus. Small changes are possible before launch.
Want this course? Tell us.
Send one message on WhatsApp, and we will tell you as soon as the launch date is fixed.