The AI connection path matters because Make's documented credit behaviour can vary by connection type and token or other usage factors. Compare the complete usage boundary, not just the number of AI steps in a scenario.
Establish what the documentation actually says
Make states that non-AI apps generally use one credit per operation, while AI features can vary with connection type and usage factors. Its guide also describes credits consumed by scenario-triggered features and AI-agent chat.
This does not establish one fixed rate for every AI request. Nor does it establish which provider costs, model choices or access arrangements apply to your proposed setup. Those require current, feature-specific checks.
Write two connection-path cards
Practical editorial advice: prepare one card for a provider-managed path and another for a custom provider connection, where the relevant feature supports them. The labels organise your research; they do not assert that both paths exist for every feature.
On each card, answer:
- Connection owner: who controls the connection and can revoke it?
- Usage rule: what does current Make documentation say for this feature and path?
- Additional billing boundary: are there separate provider terms or charges to investigate?
- Measurement basis: operations, tokens, another factor, or a combination?
- Visibility: what evidence will let you reconcile usage?
- Failure owner: who investigates rejected or interrupted requests?
Mark unanswered fields “unknown”. Do not turn an unknown into a zero-cost assumption.
Compare a bounded hypothetical workload
Suppose an Indian internal research team plans to summarise 200 synthetic notices monthly. Each notice contains approximately 1,500 words, with a requested summary of 120 words. No customer or confidential material is used.
Evaluate both connection cards against the same notices and requested outputs. Record actual input and output usage where available, completion failures and repeat attempts. Word counts are payload descriptors, not token estimates or credit quotes.
If one path requires repeated attempts to meet the summary specification, that affects the workload comparison. Assess usable completed summaries rather than comparing only successful single requests. Any rupee budget should identify unresolved currency and billing assumptions instead of presenting false precision.
Make the access decision last
Stop if you cannot identify the applicable usage rule or the party responsible for each potential charge. Resolve those questions before scaling the workflow.
If exploring marketplace access, check the automation catalog for access format and save the exact terms before buying. Account, code or token offers are not proof of official subscription equivalence. Marketplace terms cannot substitute for Make's feature documentation or establish an AI provider's billing policy.
Sources and scope
Source material checked 29 October 2026; features/policies may change. Make: Credits establishes generally different non-AI and variable AI credit behaviour. The comparison cards and workload are editorial advice and a hypothetical example, not confirmed connection availability, pricing or performance.