Make

What it is

Make (formerly Integromat) is a managed visual workflow automation platform. Its scenarios expose routers, iterators, aggregators, and error handlers on a canvas, making branching and bundle processing visible in the workflow model.

It runs as a managed cloud service. Make bills in credits: most non-AI module operations use credits on an operation basis, while some built-in AI features can also vary with token or other usage. Model a representative scenario instead of treating every workflow as a fixed operation count.

Who it’s for

Make is a candidate for operations teams that need a managed visual builder with explicit branching and iteration. Evaluate it with the people who will maintain the scenario and with representative bundles, errors, and retries.

It is a poor fit when self-hosting is required or when most of the workflow belongs in application code. Those constraints justify comparing n8n, Pipedream, or Windmill.

Strengths

• Visual model. Routers, iterators, aggregators, and error handlers are visible objects on the canvas. Compared with Zapier’s nested paths, Make workflows are easier to hand off to a teammate three months later.
• Integration depth. Make publishes a broad app catalog, but the useful unit is whether each required trigger and action exists.
• Credit billing. Most non-AI operations map to credits; built-in AI and other features can use different credit rules.
• Template gallery. A real library, not the polite Zapier “Top Zaps” page. Useful for cold-starting a workflow.

Weaknesses / Watch out

• Managed-only deployment. Make does not offer self-hosting, but that is not the same as having no regional choice. At organization creation, Make documents US and EU data-center locations for storage and processing, and says the selection cannot be changed later. Region selection can help with placement requirements; it does not by itself satisfy every compliance obligation or provide on-premises runtime control.
• Credit consumption. Iterators, aggregators, routers, and retries can multiply module operations and therefore credit use. Inspect scenario history with real data.
• Vendor ownership. Make is managed-only, so deployment and roadmap control stay with the vendor.
• Lock-in. Scenarios export to a JSON blueprint, but module logic is Make-specific; migrating to n8n or Activepieces means redoing the work.

Best paired with

• The CRM in the real workflow after verifying the exact triggers, actions, fields, and rate limits required.
• OpenAI or Anthropic via the dedicated AI modules — competent but not as flexible as wiring them into n8n’s AI nodes.
• n8n as a comparison tool — if you find yourself reaching for Make’s code modules a lot, the migration math may favor n8n.

Pricing reality check

Make replaced “operations” with credits as its billing unit. Most non-AI module runs still map to credits, while Make’s built-in AI features can have dynamic token-based credit use. Check the current pricing page, then replay a representative scenario to measure credits before committing to a plan; iterators, aggregators, data operations, and retries can change the total.

Frontier verification — price the scenario, not the diagram

Make documents credits as its usage unit, with rules that can vary by module and product feature. A four-module scenario is not automatically four credits once iterators, repeated bundles, AI usage, retries, or aggregators are involved.

Before choosing a plan, run a representative scenario and record the credit total at realistic volume. Keep the result date-stamped because billing rules are a volatile product surface.

Current sources: Make credits documentation, Make pricing, and Make organizations documentation for the US/EU data-center choice and its creation-time constraint. Regional-hosting wording verified 2026-09-20.

Frontier verification — agentic work stays inside the scenario boundary

Make is integrating AI and agent capabilities with the existing scenario canvas rather than replacing visual automation. Evaluate the actual modules, tool permissions, execution log, human review point, error handler, and recovery path; do not infer maturity from the “agent” label alone.

Keep deterministic routing, validation, and external writes explicit when they do not need model judgment. Use an agentic step only where ambiguity or iterative tool use creates decision value, then test failure and retry behavior with representative bundles.

Verdict

Conditionally recommended. Make fits teams that value a managed visual canvas and need explicit branching or bundle processing. Compare n8n when self-hosting or code control matters, and compare Zapier when the deciding factor is a specific managed connector and a simpler operating model.

Sources

• Official site & pricing: https://www.make.com and https://www.make.com/en/pricing
• Operations explained: Make operations documentation
• Organization data-center locations: https://help.make.com/organizations