Make provides a visual way to connect apps and automate work. It helps users build processes with clear inputs, branches and outputs, including workflows…
Make
Explore features, practical uses and pricing below.
Make is a visual automation platform for connecting applications, transforming data and coordinating multi-step processes. Users assemble scenarios from triggers and action modules, map information between them and control which paths run. It suits operations teams, automation specialists and developers who can define a repeatable business process and inspect the data passed through it. AI steps and agents can be added where a task needs language processing or flexible judgment.
A router splits a scenario into routes so the same input can be processed in different ways. Filters allow only data that meets specified conditions to continue. For example, every new prospect could be recorded in a CRM, while only prospects meeting a qualification rule trigger a staff notification.
The scenario tools also include inputs and outputs, variables, notes and reusable subscenarios. These help turn an expanding workflow into smaller understandable parts. Mapping must still match the connected application's fields and expected data types; a visual connection alone does not establish that the destination receives the right information.
Make's AI Agent (New) app is documented as open beta. It combines instructions, knowledge, input and tools inside a scenario. Tools can include modules, scenarios and MCP tools, letting an agent use selected application actions while handling variable requests.
The documentation distinguishes three approaches: fixed scenarios for predefined logic, AI app steps for bounded tasks such as summarization, and agents for flexible decisions. That distinction is useful when designing a process. A simple record synchronization may need explicit field rules, while categorizing an unfamiliar support message may benefit from an AI step. An agent's output still needs validation before it controls consequential actions.
Start with a test support form. Map the request ID, email and message into a ticketing system, then route urgent requests to a staff notification. Add a bounded AI classification step if messages need categorization, and send uncertain cases for review. Test missing fields, duplicate submissions and a disconnected destination service. Inspect the input and output of each module before turning on scheduled or event-triggered execution.
Error handlers can skip a failed bundle, retry work or supply a substitute value, depending on the chosen handler. Incomplete executions store unfinished runs for retry or manual resolution when enabled; the documentation says this setting is disabled by default. Decide explicitly which failures should be retried and which need staff attention.
Connected services impose their own permissions, schemas and rate limits. Retrying a workflow that already completed an external action can create duplicate records or notifications unless the process accounts for that. AI-agent behavior can also vary, and beta functionality and pricing may change. Keep a clear boundary between an agent's recommendation and the actions its tools are allowed to take.
Make has a Free plan and paid tiers with different credit allowances, scheduling and team features. Check the current plan comparison for the required execution frequency and administration controls. The AI-agent credit guide distinguishes Make's AI Provider from custom provider connections: operations, called tools and token usage can affect cost, and a custom provider can bill separately. Estimate usage from the complete scenario rather than counting only incoming events.