Building AI Agents
Description
In this hands-on training, participants will master the creation of autonomous, no-code AI agents using natural language prompting within the LangSmith environment. It covers the design of ambient workflows, the implementation of “Human-in-the-Loop” governance for safety, and the utilization of advanced architectures like Sub-agents. By the end of it, trainees will be able to construct, govern, and deploy secure, intelligent agent systems that seamlessly integrate with enterprise applications to automate complex professional tasks.
Indicative Duration: 12 training hours
*Duration is adjusted based on the final scope and the target audience.
Scope
| 1. Environment Setup & Architecture Overview | – Account Provisioning: – API Key Integration – Interface Tour |
|
| 2. Prompt-to-Agent Construction | – The Natural Language Build – Clarification Loop – Authentication |
|
| 3. Analyzing the Generated “Brain” | – Instruction Inspection – Toolbox Verification – First Run & Trace |
|
| 4. Tool Expansion & Logic Handling | – Adding Calendar Capabilities – Conflict Resolution Logic – Manual Instruction Tuning |
|
| 5. Human-in-the-Loop (HITL) Governance | – Configuring Approval Gates – The Approval Workflow |
|
| 6. Ambient Agents & Triggers | – Setting up Triggers – Managing the Agent Inbox – Status Visualization |
|
| 7. Multi-Agent Architecture (Sub-agents) | – Decomposition Strategy: – Sub-agent Configuration – Orchestration |
|
| 8. The Model Context Protocol (MCP) | – Connecting External Apps – Data Transformation Pipeline – Prompting for Schema |
|
| 9. Domain Knowledge (Skills) | – Creating Custom Skills – File System Exploration – Final Integration Test |
|
Learning Objectives
Upon completion of the course participants will be able to:
- Set up and configure an AI agent environment with secure API access
- Build and refine prompt-driven agents using structured instruction patterns
- Analyze agent behavior and validate tool interactions through tracing mechanisms
- Implement governance mechanisms such as approval gates and Human-in-the-Loop controls
- Understand multi-agent orchestration patterns and MCP-based integrations
- Extend agents with custom skills and test complete end-to-end workflows
Target Audience
- Roles: Backend Engineers, Software Architects, Technical Leads, Solution Architects
- Seniority: Mid-Level to Senior Professionals
Prerequisite Knowledge
- Basic familiarity with how Large Language Models (like ChatGPT) operate is helpful but not mandatory
Delivery Method
Sessions can be delivered via the following formats:
- Live Online – Interactive virtual sessions via video conferencing
- On-Site – At your organization’s premises
- In-Person – At Code.Hub’s training center
- Hybrid – A combination of online and in-person sessions
The training methodology combines presentations, live demonstrations, hands-on exercises and interactive discussions to ensure participants actively practice AI in realistic work scenarios.
