Make is a visual automation platform for connecting apps, APIs, databases, and AI models. Its workflows are called scenarios and are built by joining modules on an open canvas.
For this review, we assessed the scenario builder, data bundles, routers, error handling, AI Agents, testing tools, and credit pricing.
The canvas is Make’s defining feature. Each module represents an action. Lines show how data passes between those actions. Routers divide the workflow into conditional branches. Filters decide which records can continue.
This becomes valuable when a process has several outcomes. An incoming order might need separate routes for payment failure, low stock, international delivery, and manual review. Make shows those routes on the same map. You can see where a record entered and which condition it passed.
The harder concept is the bundle. Make passes data between modules as structured bundles. If one module returns 50 records, the next module may run 50 times. Add an iterator and those records can split again.
This is where credit use can grow unexpectedly. A scenario with six visible modules does not necessarily cost six credits. If an early module produces 100 bundles, later modules may run once for every bundle. The scenario can consume hundreds of credits from one trigger.
Aggregators help control this. They collect several bundles and rebuild them into one array, table, or block of text. This is especially useful before an AI step. Sending one combined digest is usually easier to control than making the model process every record separately.
Make’s error handlers are another strong feature. You can attach a recovery route to the module most likely to fail. The scenario can retry later, ignore one bad record, use substitute data, or stop while preserving completed database changes. That level of control is useful when an occasional API failure should not cancel an entire process.
Testing has improved. You can replay trigger data from an earlier run against the current scenario. This means you do not always need to create another live order or form response. Replays still consume credits, but they make difficult triggers much easier to test.
AI Agents can use a model, knowledge, and connected tools inside a scenario. Their cost depends on the AI provider. Make’s provider charges for the operation and the model tokens. Paid plans can connect an outside provider account, leaving Make to charge for the module and tool runs while model usage is paid separately.
The credit system therefore needs planning. Most ordinary module actions cost one credit. Polling triggers can also consume credits when they check for new data. AI features and large bundle counts can cost more. Extra credits carry a 25% premium, so inefficient scenarios become expensive before they become obviously broken.
