Make connects apps through visual cloud workflows called scenarios. Its AI Agent (New) adds a step that can interpret changing inputs and choose among the tools you give it. You supply instructions, a model provider and any reference knowledge; each run adds the request it must handle. This is useful when an inquiry needs judgment before following an otherwise predictable route. Keep known rules in ordinary scenario steps, and use a bounded AI module for a simple transformation such as summarizing text. The agent app is in open beta. Make's provider is available across plans, while custom provider connections require paid access; functionality and pricing may change.
In this profile
When it fits
Useful for
- Adding a judgment step to a visual cloud automation scenario
- Choosing among fixed logic, a bounded AI transformation and an agent inside one platform
Limits to consider
- Adding an agent where the rule is already known and explicit
- Assuming open-beta behavior or pricing will remain fixed
Capabilities and platforms
- Agents accept configured instructions, provider, knowledge and scenario inputs
- Agent tools can use modules, scenarios, MCP connections and other agents
- Ordinary scenarios and AI modules provide alternatives to agent reasoning for fixed or bounded work
Where it runs
- Cloud platform
Access and setup
Make account; the documented Make provider spans plans, while custom provider connections require paid plans. New-agent functionality and pricing remain subject to beta changes.
Current access and pricing- Open the current Make AI Agent (New) documentation and check beta availability and provider access for the account.
- Configure instructions and a small synthetic input, selecting only the modules or scenarios required as tools.
- Test the agent inside a scenario and inspect its proposed actions and output; keep fixed rules in explicit scenario logic.
Proposed exercise · not hands-on tested
Draft an internal inquiry route
Input
Synthetic inquiries and a known set of destinations, including an unresolved option.
- List the allowed destinations and apply known eligibility rules with ordinary scenario logic. Give the agent only the variable part: interpret the remaining synthetic inquiry and propose a destination.
- Ask for a destination and short reason, with an explicit unresolved result when the inquiry does not support a choice. Keep reference knowledge separate from the new inquiry so the scenario has a stable routing brief.
- Inspect the scenario's run history and any tool actions. Compare the proposed destination with the known test labels before connecting an external send step or enabling recurring execution.
Artifact to inspect
A routing draft with fixed rules separated from judgment.
What to verify
This is an unexecuted exercise. Check that fixed rules remain fixed, unresolved inquiries reach review and the agent uses only the tools assigned to its routing role. Compare the reason as well as the label so an accidentally correct destination does not hide a faulty interpretation.
Data handling and permissions
Make scenarios run in the cloud, and their connected services and model providers determine additional data destinations. A custom-provider connection changes the processing path, rather than making the scenario local. Review what each module, tool and provider receives before passing internal records through an agent. The open-beta and provider-access descriptions do not establish a blanket training, retention or compliance guarantee. Check current terms, connection permissions and scenario records for the account being used. The proposed routing exercise uses synthetic inquiries so its result can be evaluated before real messages or customer material enter the workflow. Keep publishing and recurring execution separate from inspecting a draft.
Other options
- n8n
- Consider it when self-hosting or more explicit technical operation matters.
- Zapier Agents / AI by Zapier
- Compare for agentic tool use alongside ordinary Zap steps.
Sources and scope
Documentation-based product profile; no hands-on testing. The proposed workflow is an authored exercise, not a verified outcome or a practical guide. Capabilities and policy statements describe vendor documentation checked October 2, 2026; task-fit comparisons are editorial judgments. Plan, region, administrator and interface limits may apply.
Source review: . Product access and terms can change.
- 01Current app, beta status, providers and toolshelp.make.com
- 02Official plan and pricing informationwww.make.com