🔥 Feature Spotlight: Make AI Sub-Agents

[Valery.Mezencev]

Make AI Agents can now delegate tasks to other Make AI Sub-Agents, enabling agent orchestration inside your scenarios.

Hey Makers As AI workflows grow (more decisions, more tools, more edge cases), stuffing everything into one Make AI Agent starts to break down. Too much context, too many instructions, too hard to debug when something goes wrong.

That’s why we are excited to introduce Make AI Sub-Agents, a new agent orchestration capability in Make: you can now add Make AI Agents as tools inside other Make AI Agents.

One Make AI Agent receives the task, selects the appropriate specialist, and returns a final result. No hard-coded handoffs needed.

What’s new with Make AI Sub-Agents

• Agent orchestration in real time: The orchestrator Make AI Agent calls the right specialist Make AI Agent based on what the request needs, instead of following one fixed path every time.
• Build focused AI agent specialists: Give each Make AI Agent its own tools, system prompt, and job, like CRM lookup, refund handling, or invoice review.
• Test and improve Make AI Agents one by one: You can tune a refund Make AI Agent without touching your account lookup Make AI Agent, then plug both back into the same workflow.
• Reuse proven Make AI Agents across scenarios: Build a specialist Make AI Agent once, then use it in multiple orchestrators across support, finance, and operations.

What you can automate with Make AI Sub-Agents

• Customer support flows: An orchestrator Make AI Agent reads the customer message, sends account questions to a lookup Make AI Agent, refund requests to a refund Make AI Agent, and edge cases to a human handoff agent.
• Finance operations: One Make AI Agent handles incoming requests, then routes document checks, invoice matching, and approval prep to separate specialist Make AI Agents before returning one clear result.
• Internal service desks: Route employee requests across access checks, policy answers, and ticket creation, while keeping one AI point of contact for the whole exchange.

How to create Make AI Sub-Agents

Here is how it works:

1. Add a Make AI Agent to your scenario. This will be your AI agent orchestrator.
2. Inside that Make AI Agent orchestrator, click Add tools and find Make AI Agents in the tools list.
3. Configure the Make AI Agent as a tool and click Add. It will link to your orchestrator as a Sub-Agent.
4. Repeat to add as many Sub-Agents as you need, each one a focused specialist.

One thing to note: currently, Sub-Agents support one level of nesting. You can add Make AI Agents as tools inside an orchestrator Make AI Agent, but you cannot add Sub-Agents to Sub-Agents (yet).

That’s agent orchestration, Make-style: one orchestrator, many specialists, one result. What would you build first?

Availability

Make AI Sub-Agents are available on all Make plans, including the Free plan. You can start building with nested Make AI Agents today, no upgrade needed.

Try Make AI Agent orchestrator with sub-agents yourself

Want to see Make AI Sub-Agents in action before building your own? Clone this example scenario and explore how the AI agent orchestrator and its agent specialists are set up: https://ma.ke/ai-sub-agents-scenario

Demo time

Happy automating!

Valery from Make

Helpful resources

Make AI Agents full guide

Learn how to add Make AI Sub-Agents

Make AI Agent best practices (token optimization, testing, data security and more)

[Vicente_Moreno]

top! truly agentic system directly in Make!

[Valery.Mezencev]

Agreed, @Vicente_Moreno ! The orchestrator + specialist setup really does feel like a big step toward true agentic workflows. Do you have any cool use cases in mind you’d want to share? Would love to hear what you end up building with it

[DG_AdSystems]

What’s the significant advantage to this, when compared against calling scenarios which themselves have Ai agents within them?

[Sarah_D]

I think this will be super useful. I’ve been chaining Agents as a way to get better outputs from not giving too much to one, but sub-agents should be able to do this more efficiently.

[Vicente_Moreno]

Customer support, lead management and follow up(this one is the unlock) - and the one I did in the video!
Anything that needs multiple steps is a good use case! You imagination is the limit !

[Temi_Josh]

Very interesting feature. For production workflows, how does the orchestrator handle failures or conflicting responses from sub-agents? Is there a way to define fallback behavior or have the orchestrator retry with another specialist?

[Valery.Mezencev]

Good question! You’re right that adding a scenario with an AI Agent inside it as a tool already gives you AI-led routing, the orchestrator decides when to call it either way.

The difference is in what you have to build to get there. A scenario-as-tool needs its own trigger and response setup, its own input/output structure, and runs as a separate scenario execution every time it’s called. A Sub-Agent is just another AI Agent you pick directly as a tool, no wrapping scenario required, and Make already knows its input/output shape since it’s a native agent-to-agent connection.

Both are valid depending on what you’re building. If you just need a focused specialist doing one job, Sub-Agents is the more direct route. If you need more control over what happens around that call, retries, custom logic, other modules in between, a scenario-as-tool still gives you that flexibility.

[Valery.Mezencev]

Love that you’ve already been solving this by hand, Sarah! That’s exactly what Make AI Sub-Agents is built for: instead of chaining agents and passing outputs yourself, you add each specialist as a tool inside one orchestrator, and it decides who to call and how to combine the results. Curious what you’re chaining today, would love to see how it maps over.

[Valery.Mezencev]

Yes, you can build this fallback behavior with Make AI Agents. Define the fallback logic in the orchestrator’s system prompt (e.g. “if the refund specialist fails or gives a low-confidence answer, try the escalation specialist instead”), and add that fallback specialist as a Sub-Agent tool so the orchestrator actually has it available to call. With both in place, the orchestrator can reason its way into retrying with the right specialist.

There isn’t a dedicated fallback/retry setting built into AI Agents itself yet, it’s achieved through prompt design plus having the right specialist on hand. If this turns out to be a common pattern, it’s a good candidate for something more native down the line.