In this intensive one-day workshop, we leave the basic chat wrappers behind and focus on the infrastructure required to run agents reliably in production. We will dissect the Agent Harnessthe surrounding system that gives an LLM its memory, tools, and guardrails. Instead of building one large cumulative project, we will tackle a new architectural pattern in every lesson. You will see a concept introduced, watch it live-coded in TypeScript/Node.js, and learn how to apply it. By the end of the day, you will understand how to use durable execution to survive network failures, sandbox tool execution, and orchestrate complex workflows across multiple specialized sub-agents.
What You Will Master
The Agent Harness: Understand the critical middleware layer that provides guardrails, observability, and context injection.
Durable Execution: Replace brittle scripts with workflow engines that checkpoint state, auto-retry tool failures, and recover from crashes.
Runtime Sandboxing: Safely execute untrusted agent code and tool calls in isolated environments.
Advanced Memory Infrastructure: Hydrate agents with exactly the right context using state stores, sliding windows, and summarization loops.
Multi-Agent Orchestration: Route intent through Supervisor nodes, implement multi-agent handoffs, and manage hierarchical agent networks.
Human-in-the-Loop (HITL): Use durable suspend/resume primitives to handle unbounded wait times for human approval.
Prerequisites
Comfortable with TypeScript and Node.js.
Basic understanding of LLM APIs and standard agent loops (receive message think act respond).
Familiarity with asynchronous programming and API integrations.
Who is this for?
This workshop is for full-stack JS/TS developers transitioning into AI systems architecture, engineers who have built basic agents and want to learn how to make them fault-tolerant and scalable, or technical leads needing to understand the orchestration and infrastructure layers of modern AI applications.

Homepage:

In this intensive one-day workshop, we leave the basic chat wrappers behind and focus on the infrastructure required to run agents reliably in production. We will dissect the Agent Harnessthe surrounding system that gives an LLM its memory, tools, and guardrails. Instead of building one large cumulative project, we will tackle a new architectural pattern in every lesson. You will see a concept introduced, watch it live-coded in TypeScript/Node.js, and learn how to apply it. By the end of the day, you will understand how to use durable execution to survive network failures, sandbox tool execution, and orchestrate complex workflows across multiple specialized sub-agents.
What You Will Master
The Agent Harness: Understand the critical middleware layer that provides guardrails, observability, and context injection.
Durable Execution: Replace brittle scripts with workflow engines that checkpoint state, auto-retry tool failures, and recover from crashes.
Runtime Sandboxing: Safely execute untrusted agent code and tool calls in isolated environments.
Advanced Memory Infrastructure: Hydrate agents with exactly the right context using state stores, sliding windows, and summarization loops.
Multi-Agent Orchestration: Route intent through Supervisor nodes, implement multi-agent handoffs, and manage hierarchical agent networks.
Human-in-the-Loop (HITL): Use durable suspend/resume primitives to handle unbounded wait times for human approval.
Prerequisites
Comfortable with TypeScript and Node.js.
Basic understanding of LLM APIs and standard agent loops (receive message think act respond).
Familiarity with asynchronous programming and API integrations.
Who is this for?
This workshop is for full-stack JS/TS developers transitioning into AI systems architecture, engineers who have built basic agents and want to learn how to make them fault-tolerant and scalable, or technical leads needing to understand the orchestration and infrastructure layers of modern AI applications.

Homepage:

In this intensive one-day workshop, we leave the basic chat wrappers behind and focus on the infrastructure required to run agents reliably in production. We will dissect the Agent Harnessthe surrounding system that gives an LLM its memory, tools, and guardrails. Instead of building one large cumulative project, we will tackle a new architectural pattern in every lesson. You will see a concept introduced, watch it live-coded in TypeScript/Node.js, and learn how to apply it. By the end of the day, you will understand how to use durable execution to survive network failures, sandbox tool execution, and orchestrate complex workflows across multiple specialized sub-agents.
What You Will Master
The Agent Harness: Understand the critical middleware layer that provides guardrails, observability, and context injection.
Durable Execution: Replace brittle scripts with workflow engines that checkpoint state, auto-retry tool failures, and recover from crashes.
Runtime Sandboxing: Safely execute untrusted agent code and tool calls in isolated environments.
Advanced Memory Infrastructure: Hydrate agents with exactly the right context using state stores, sliding windows, and summarization loops.
Multi-Agent Orchestration: Route intent through Supervisor nodes, implement multi-agent handoffs, and manage hierarchical agent networks.
Human-in-the-Loop (HITL): Use durable suspend/resume primitives to handle unbounded wait times for human approval.
Prerequisites
Comfortable with TypeScript and Node.js.
Basic understanding of LLM APIs and standard agent loops (receive message think act respond).
Familiarity with asynchronous programming and API integrations.
Who is this for?
This workshop is for full-stack JS/TS developers transitioning into AI systems architecture, engineers who have built basic agents and want to learn how to make them fault-tolerant and scalable, or technical leads needing to understand the orchestration and infrastructure layers of modern AI applications.

Homepage: