How many concurrent calls do I need, and how do I increase the limit?
The concurrent active calls per instance quota counts every call that is in progress in Amazon Connect at the same moment. That includes callers in the IVR and callers waiting in a queue, not just calls with an agent. Its default for new accounts is low (10), so it's the quota most contact centres hit first.
Working out your peak
- Take the number of calls in your busiest hour, not your daily average.
- Multiply by the average time a call spends in Amazon Connect (IVR + queue wait + talk and hold), in hours. That's the average number of calls in progress. For example, 600 calls an hour × 6.5 minutes ≈ 65 calls in progress.
- Calls don't arrive evenly, so the busiest moment is higher than the average. The calculator on this page uses the 99.9th percentile of random (Poisson) arrivals, then adds 25% headroom for the figure to request.
The same calculator runs Erlang C, the standard contact centre staffing model: it shows how many agents you need on the phones to answer, for example, 80% of calls within 20 seconds, the average wait, agent occupancy, and how many to schedule after shrinkage (breaks, training, absence). After-call work keeps an agent busy but doesn't use a concurrent-call slot.
Defaults versus your real quotas
AWS publishes default values for new accounts and notes that the applied values in your account may be lower, because defaults have changed over time. Increases you've already had approved also make them higher. For exact results, run aws service-quotas list-service-quotas --service-code connect --quota-applied-at-level ALL in your instance's region and import the output. The planner then checks against your account's real values and fills in the quota codes.
Requesting an increase
- Use the Service Quotas console, or
aws service-quotas request-service-quota-increase. Most Amazon Connect quotas are set per instance, so include--context-idwith the instance ARN. Viewing and managing per-instance quotas from the CLI needs AWS CLI 2.13.20 or later. - Quotas are per region. If you run a second region for disaster recovery or Global Resiliency, request the same increases there, or failover will hit the old limits.
- AWS says small increases can be approved within hours, larger ones can take up to 3 weeks, and very large worldwide increases can take longer. Request well before go-live.
- Some quotas, including Amazon Q in Connect (AI agents) and certain analytics evaluation limits, are changed through an AWS Support case instead. The planner writes the case text for you.
API rate limits
Amazon Connect API limits apply per account and region, and are shared by every application, instance and user. Unless AWS lists a different value, each API allows 2 requests per second with bursts of 5. Real-time metrics (GetCurrentMetricData) allow 5 per second, burst 8, so 20 wallboard screens refreshing every 5 seconds already need 4 per second. Use the API rate planner to add up every integration. Where you're close to the limit, poll less often, cache results, use event streams (Kinesis or EventBridge) instead of polling, and retry with exponential backoff.