Sub-Agents in Make: Building a Multi-Agent System Inside One Scenario

In this build, Vicente takes a real, already-live scenario — a lead-qualification flow for a TikTok management agency — and shows how the new Sub-Agents feature lets him break one overloaded AI agent into a coordinated team of specialists, all inside a single Make scenario. 🔍 Inside the Build: * The Problem with One Giant Agent: Vicente shows his original single-agent setup, where one agent tried to qualify leads, scrape TikTok data, draft an analysis, log to Airtable, and notify him — all at once, often overwhelming its context window. * Turning the Main Agent into an Orchestrator: He renames the original agent to "Agent Orchestrator" and shows how to attach other Make agents to it as tools — each one handling a single, focused task. * Building the Research Agent: A dedicated sub-agent researches the lead's company and social presence, scrapes their TikTok performance data, and returns a structured analysis — without the orchestrator needing to manage any of those steps itself. * Building the Qualification and Proposal Agents: Two more sub-agents check the research against qualification criteria and draft a tailored proposal, each with its own narrow instructions. * Watching the Handoffs in Real Time: Vicente reruns a test lead and walks through the execution log step-by-step — showing exactly how the orchestrator delegates to each sub-agent, waits for results, and strings the outputs together to trigger a phone notification and send a final email. This is a strong watch for anyone whose Make agent has hit a wall from doing too much at once — Vicente shows how breaking work into sub-agents makes testing, debugging, and reusing each piece far easier. 🛠️ Featured Stack: * Make AI Agents (Orchestrator + Sub-Agents as Tools) * Web Scraping Module (Company & Social Research) * Airtable (CRM Record Logging) * Email & Push Notification Modules (Final Delivery) 💡 Pro-Tip from Vicente: "Once one part breaks, the whole agent doesn't die." Splitting a single overloaded agent into sub-agents means you can test and fix one piece at a time — and copy a working sub-agent into other scenarios once it's solid.