In this video, I look at the difference between single-user agents like OpenAI and systems that people are building for themselves versus multi-user agents, which are built for production with upwards of thousands of users. 

For more tutorials on using LLMs and building agents, check out my Patreon
Patreon: https://www.patreon.com/SamWitteveen
Twitter: https://x.com/Sam_Witteveen 

🕵️ Interested in building LLM Agents? Fill out the form below
Building LLM Agents Form: https://drp.li/dIMes

👨‍💻Github:
https://github.com/samwit/llm-tutorials

⏱️Time Stamps:
00:00 Intro
00:41 Single User Agent vs Multi User Agent
05:19 Agent Core vs Agent Harness
06:15 What Breaks When You Go Multi User?
09:31 What's Next

Sam Witteveen AI is a publication offering insights, tutorials, and resources for artificial intelligence (AI) enthusiasts and practitioners. Readers can learn about machine learning algorithms, deep learning frameworks, and AI applications. With tutorials, case studies, and expert interviews, Sam Witteveen AI provides  guidance and expertise for building and deploying AI solutions.

Sam Witteveen

Building an AI agent demo is very different from deploying it to many users. Single-user agents can rely on simple state (markdown files), tolerate inefficiency, and skip guard rails. Multi-user agents require state isolation to prevent context leakage between users, cost controls and quotas to prevent runaway spending, latency management with queues and retries, observability and audit logging, and security measures against prompt injection and tool misuse. The core distinction is framed as 'agent core' (planning, memory, tools) vs. 'agent harness' (isolation, auth, cost controls, observability) — single-user systems are core-heavy while multi-user systems must nail the harness first.

Single-User vs Multi-User Agents: What Actually Changes