{{ summaryTitle }} {{ summaryValue }} {{ summaryLine }} {{ badge.label }} {{ badge.value }}
Start with a representative workload or keep Custom for the values already shown.
Choose the deployed API family; the visible inputs and billing meters update with it.
Use the published AWS US examples or an explicit planning multiplier; all output remains USD.
Enter 1.00 for the base example rates or a verified multiplier for this planning scenario.
×
Enter the observed or forecast monthly traffic and its scale.
{{ requestSizeHelp }}
KB
Use the average response payload after application-level compression.
KB
Select streaming only when the REST integration uses response transfer mode STREAM.
Cache adds an hourly charge; client API calls remain billable even on cache hits.
Use an observed hit ratio when available.
%
PrivateLink endpoint hours are charged for each provisioned Availability Zone.
AZs
Enable only when the account and billing month qualify for AWS's time-limited offer.
{{ include_free_tier ? 'Enabled' : 'Disabled' }}
Enter 0 to omit budget variance from the decision review.
$
Estimate withheld. {{ workflowFeedback }}
Add client retries, webhook replays, probes, or test leakage that is absent from the entered traffic.
%
Subtract an account-specific allowance before the public response-transfer rate.
GB
Off uses the cited AWS examples. Turn on only after checking the target Region, account terms, and billing date.
{{ use_custom_rates ? 'Enabled' : 'Disabled' }}
Leave 0 unless managed API Gateway Portals are in the project scope.
portals
Enter products beyond the first 10 included with each portal.
products
Cost componentMonthly costBilling basisCopy
{{ row.label }}{{ row.value }}{{ row.detail }}
MeterQuantityApplied ruleCopy
{{ row.label }}{{ row.value }}{{ row.detail }}
{{ item.title }}

{{ item.body }}

Cloud Egress Cost Calculator Estimate cloud egress spend from measured traffic and a marginal rate card with verified reductions, peak-month stress testing, and budget variance.
AWS Lambda Cost Calculator Estimate monthly AWS Lambda Functions cost across requests, GB-seconds, provisioned capacity and optional meters with editable regional rates.
Object Storage Cost Calculator Estimate monthly object storage cost from capacity and request volume with retrieval, egress, lifecycle and editable provider rate assumptions.
Kafka Partition Count Calculator Plan a Kafka topic partition count from throughput and consumer parallelism, then check broker spread and keying risk before rollout.

API Gateway cost begins with traffic, but raw call count is rarely the whole bill. The API family determines the primary meter. Payload size can turn one client action into several billable units, and some designs add response transfer, connection minutes, cache hours or private endpoint charges.

HTTP APIs suit many lightweight request-response workloads. REST APIs offer a different feature and pricing profile. Private REST APIs add interface endpoint cost to keep access within private networks, while WebSocket APIs charge for messages and the time connections stay open. The cheapest family on paper may not provide the behavior an application needs.

Raw traffic
Client requests or messages before retries, replays and billing-size increments.
Metered units
Traffic after overhead and payload chunking have been applied.
Tiered charge
Each traffic band is priced separately rather than applying the final tier's rate to every unit.
Secondary meters
Transfer, connection minutes, cache capacity, private endpoints and portal add-ons when applicable.
API familyPrimary meterSize incrementImportant extra
HTTP APIRequests512 KB request chunksPublic response transfer
REST APIRequestsOne unit normally; streamed responses use 10 MB incrementsOptional cache and response transfer
Private REST APIRequestsOne unit normallyInterface endpoint hours and processed data
WebSocket APIMessages32 KB message chunksConnection minutes

Average payload size is a planning shortcut. Averages can hide a small number of large uploads or messages that cross a chunk boundary, so production billing data or a payload-size distribution gives a stronger forecast. Retry storms, webhook replays and health traffic can also make metered volume exceed business transaction count.

A gateway estimate is not a full application estimate. Backend compute, Lambda duration, databases, logs, WAF, NAT gateways, custom domains, support and other AWS services remain separate. Model those charges beside the gateway line before using the result as an application budget.

How to Use This Tool:

Start with observed or forecast traffic, then select the API family and pricing assumptions that match the deployment.

  1. Choose a Workload preset or keep the custom values, then set API type.
  2. Choose the Rate profile. Use a custom multiplier or custom rates only when they come from the intended AWS Region or agreement.
  3. Enter monthly requests or messages in the correct scale and provide the relevant average request or response size. For WebSocket traffic, add average active connections and connected hours per day.
  4. Add retry overhead, included response transfer and any eligible free-tier units. For REST workloads, include cache only when that cache is actually provisioned.
  5. For Private REST APIs, enter the number of interface endpoint Availability Zones. Add portal counts only when those paid features are part of the plan.
  6. Compare Cost breakdown with Metering ledger. If the estimate is withheld, correct the first invalid traffic, payload, rate or connection value before using the total.

Interpreting Results:

Estimated monthly gateway cost is the sum of the active gateway cost components in USD. Effective cost per million divides that total by metered units, so it can rise when fixed cache, endpoint or portal charges are spread across modest traffic.

  • Use Metering ledger to check raw traffic, overhead, chunk factor, allowances and paid units before trusting the cost total.
  • A positive Budget variance means the estimate is over the entered monthly budget; a negative value means it is under.
  • Cache hits represent backend requests avoided. Client requests that pass through API Gateway remain billable in this model.
  • The API comparison holds the workload assumptions constant. It compares cost shapes, not feature compatibility or migration effort.

Treat the result as a planning snapshot tied to the entered rates. AWS prices, free-tier eligibility and transfer treatment can change by Region, account and billing arrangement.

Technical Details:

The calculation expands the selected traffic scale, adds retry or replay overhead and rounds payloads up to the billing increment for the chosen family. Any enabled allowance is subtracted after chunking, never from raw business transactions.

Formula Core:

The common traffic path is:

Toverhead=Traw×(1+o100) Tmetered=Toverhead×k Tpaid=max(0,Tmetered−Tallowance)

o is retry overhead percent. The chunk factor k is the ceiling of request KB divided by 512 for HTTP APIs, message KB divided by 32 for WebSocket APIs, or streamed response KB divided by 10,240 for REST response streaming. Other REST paths use one unit per call.

Tiered request or message cost sums the units in each tier:

Crequest=∑inTi1000000×ri×m

Ti and ri are units and USD-per-million rate in tier i; m is the selected regional multiplier. The built-in planning card uses these rate tiers:

FamilyTierRate per million
HTTP APIFirst 300 million$1.00
HTTP APIOver 300 million$0.90
REST APIFirst 333 million$3.50
REST APINext 667 million$2.80
REST APINext 19 billion$2.38
REST APIOver 20 billion$1.51
WebSocket APIFirst 1 billion messages$1.00
WebSocket APIOver 1 billion messages$0.80
ItemBuilt-in value
WebSocket connection minutes$0.25 per million minutes
Public response transfer$0.09 per GB
REST cache example$0.038 per hour for 730 hours
Private endpoint$0.01 per AZ-hour plus $0.01 per GB
Developer portal$125 per portal plus $12.50 per product monthly
Rate profiles1.00 standard; 1.21 regional premium; 1.34 high premium

Public HTTP and REST response transfer is response bytes divided by 1,048,576 KB per GB, less the entered included GB, multiplied by the transfer rate. Private REST data uses request plus response bytes and adds endpoint hours for each Availability Zone. WebSocket connection minutes equal average connections multiplied by connected hours per day, 60 minutes and a 30-day month.

The built-in allowance removes up to one million request or message units and, for WebSocket APIs, 750,000 connection minutes. It is applied only when enabled and does not establish account eligibility. Calculations retain full precision; displayed USD values are formatted for reading.

Accuracy Notes:

Use current AWS pricing for the intended Region and account. The standard profile mirrors published U.S. examples captured by the model, while the premium multipliers are planning assumptions rather than quoted regional prices. No workload values are transmitted; the estimate is calculated in the browser.

Worked Examples:

HTTP traffic crossing the first tier

At 432 million HTTP requests with small payloads, no allowance and the standard rate card, the first 300 million cost $300 and the remaining 132 million cost $118.80. The request total is $418.80 before any response transfer or add-ons.

Payload boundary

An average 513 KB HTTP request uses two 512 KB billing chunks, so one million raw calls become two million metered units. Checking payload distribution matters when many requests sit close to a boundary.

References: