Agents in Production

Overview

An agent that works in a demo is not ready for production. In production it has many users with different access rights, real load, real costs, and people who try to misuse it. This module looks at what the agent needs around it to run in production, and at the decisions that are expensive to change later. In the hands-on part you load test an agent and find where the system breaks.

What you'll cover

• ▸What an enterprise agent system is made of, and what to build vs. buy
• ▸The production checklist: the questions to ask before going live, and which module covers each one in depth
• ▸Access and identity: who can see what, and whether the agent acts as itself or on behalf of the user
• ▸Scale, cost and security: one message fanning out into many calls, load testing, guardrails, and the fronts you have to protect
• ▸Harness vs. sandbox: where an agent's tool calls run, and how to contain them

What you'll be able to do

• ✓Map a use case onto the enterprise components and decide what to build and what to buy
• ✓Load test an agent before launch and find where the system breaks
• ✓Control what an agent can see and do, with permissions enforced outside the LLM
• ✓Monitor and limit what an agent costs