Manage Multi Agents with Paperclip

Running several AI agents as a coordinated team raises the same questions as running a team of people: who owns what, who approves it, and what it costs. Dr. Alfred Ang has been answering those questions in working systems. Dr. Ang is CTO of Tertiary Infotech Academy Pte Ltd in Singapore, the training organisation behind Tertiary Courses, and a full-stack AI developer, Agentic AI solution architect and corporate trainer who builds intelligent agents and business solutions end to end, from front end and back end to databases, cloud deployment and native mobile apps. Relevant builds include an ERP prototype in which AI agents handle purchase orders and replenishment while humans approve each step through email, Asana or Telegram; an automated video production pipeline in which a web form triggers an agent that researches, scripts, renders, verifies and publishes; and an AI-powered content management system on the Claude Agent SDK with a weekly auto-blog scheduler and a self-improving chatbot. Dr. Ang also maintains Claude Code skills, works daily with coding agents such as Claude Code and Codex, and created an AI coaching simulator for redesigning jobs around accountable ownership of AI agents. An n8n HR portal adds human-in-the-loop manager approval and a guarded chatbot. The public portfolio at github.com/alfredang holds more than 300 repositories, and the common thread is governance: agents deliver real value only when goals, budgets, approvals and audit trails are designed in from the very start.

This course applies that thinking to Paperclip, a platform for organising and governing multi-agent teams. Learners start with an introduction to multi-agent AI and the platform, then install and self-host Paperclip so they understand exactly where their agents and data live. Next, they build an agent org chart with roles, titles and reporting lines, and connect agent runtimes such as Claude, Codex, Gemini and Cursor, comparing which runtime suits which kind of work. The second topic moves from setup to operations: aligning agent work to business goals and missions, managing tasks with tickets and heartbeat scheduling, controlling costs with per-agent budgets and spend tracking, and governing the whole system through audit logs, approvals, overrides and, when necessary, agent termination. Dr. Ang brings lessons from the approval workflows and scheduled agents in the portfolio, so discussions cover practical issues such as how much autonomy to grant, where a human checkpoint belongs and how to notice when an agent is quietly drifting from its brief. The teaching is hands-on throughout: learners stand up a small agent company of their own, assign it a realistic mission, watch it work through tickets and adjust budgets and permissions as they go. Everyone leaves with a running Paperclip setup and a governance checklist for taking multi-agent work into production. Time is also given to comparing results across runtimes and reading the audit trail to understand exactly why each agent made the decisions it did. To see the agentic systems that inform this course, explore Dr. Ang's public portfolio at https://github.com/alfredang