This bootcamp is a hands-on path from inference to production-style agents. You will call NVIDIA® NIM™ from cloud and local endpoints, expose and consume capabilities with the Model Context Protocol (MCP) (including a low-level server implementation), and orchestrate reasoning and tool use with LangGraph. You will then use NeMo Agent Toolkit (NAT) to connect MCP tools to NIM with YAML workflow configuration—plus observability and evaluation—before tying the stack together in a final Challenge.
This content contains 5 Labs, plus a challenge and a bonus challenge:
- Lab 1: Using NVIDIA NIM via Cloud and Local Endpoints
- Lab 2: Introduction to Model Context Protocol (MCP)
- Lab 3: Low-Level MCP Server Implementation
- Lab 4: Building Agentic Workflows with LangGraph
- Lab 5: NeMo Agent Toolkit (NAT)
- Lab 6: Challenge
- Lab 7: Bonus Challenge
The tools and frameworks used in this bootcamp are as follows
Participant:
- Knowledge in at least 1 agentic framework - e.g. LangChain/LangGraph, LlamaIndex.
- Experience in using docker to deploy workloads.
Development environment:
- Python 3.13
- UV python package manager
- Modern IDE - e.g. VSCode, Cursor, Zed
- Permission to install software and python packages
- Unix-like operating system - e.g. ubuntu, MacOS (preferred)
Miscellaneous:
- Nvidia NGC API Key
Data Scientists, Developers
To deploy the Labs, please refer to the deployment guide presented here
The duration of the tutorial is 2 hours.
The duration of the challenge is 3 hours.