Real-Time Context Engine Billing
The Real-Time Context Engine and lightning queries are fully managed Confluent Cloud capabilities. Applications, users, and AI agents use the Real-Time Context Engine to query live, continuously updated Apache Kafka® data through Model Context Protocol (MCP) tools.
Lightning queries let you query that same data directly through an API. You don’t need to build or maintain separate serving infrastructure.
Note
Direct API access, issuing a lightning query yourself without an AI agent, is an Early Access Program feature in Confluent Cloud.
An Early Access feature is a component of Confluent Cloud introduced to gain feedback. This feature should be used only for evaluation and non-production testing purposes or to provide feedback to Confluent, particularly as it becomes more widely available in follow-on preview editions.
Early Access Program features are intended for evaluation use in development and testing environments only, and not for production use. Early Access Program features are provided: (a) without support; (b) “AS IS”; and (c) without indemnification, warranty, or condition of any kind. No service level commitment will apply to Early Access Program features. Early Access Program features are considered to be a Proof of Concept as defined in the Confluent Cloud Terms of Service. Confluent may discontinue providing preview releases of the Early Access Program features at any time in Confluent’s sole discretion.
To participate in the Early Access Program, see the sign-up form.
You’re billed for using the Real-Time Context Engine and lightning queries in two ways. You pay for the compute used to materialize and serve queries, and for the storage your materialized data consumes. You don’t need to size a cluster or provision and scale anything manually. You enable the Real-Time Context Engine on a topic, and your cost follows your usage.
How CCUs are calculated
A Confluent Compute Unit (CCU) is a logical unit of processing power. It reflects the compute resources used to read from Kafka and materialize your data, making it queryable. It also reflects the resources used to serve your queries.
Usage scales with your workload. The more data you materialize or query, the more CCUs you consume, and usage scales back down when activity drops. You pay only for what you use, metered every minute.
Definition of one CCU
A single CCU covers the following:
Hot memory size: Compute for a hot memory size supporting roughly 40 GB of base storage.
Query scan: Up to 80 GB scanned by your queries per hour, whether that’s one large query or many small ones.
Ingest throughput: Up to 10 MBps of sustained data streaming in.
You pay for whichever meter is highest in a given minute. A CCU isn’t “40 GB and 80 GB and 10 MBps” all at once. It’s whichever single threshold your workload reaches.
Billing ties to familiar things about your workload, such as how much data you’re materializing, throughput, and how much your queries scan, rather than the underlying CPU or machine size, so you can reason about cost before you build against it.
Pricing
These are list prices before the regional multiplier is applied.
- Storage
Costs $0.029 per GB-month, billed hourly, based on the average monthly bytes of materialized data stored.
- Compute
Compute is measured in CCUs. The price is $0.105 per CCU-hour, which is $0.00175 per CCU-minute. Usage is metered every minute and rolled up into hourly billing increments.
Note
Confluent Compute Units (CCUs) aren’t to be confused with a Confluent Consumption Unit (also abbreviated CCU). That’s the unit used for Confluent Cloud marketplace billing, and the two are unrelated.
Prices vary by cloud region.
Cost behavior in common scenarios
If a table has no new data coming in and no queries running against it, you’re not charged any CCUs; you pay only for storage on whatever’s already materialized.
If data is streaming in but no queries have run yet, you’re still charged for the compute required to keep the table materialized and current. This continues for as long as the topic stays actively streaming, even without any queries running against it.
You’re charged for the total amount of data scanned in a minute, not the number of queries that produced it: ten small queries and one large query cost the same if they scan the same amount of data. Sustaining an unusually high query rate for a full minute can add a small additional charge, though this is rare and doesn’t affect typical usage.
If both throughput and queries stop, CCUs scale down to zero: with no data streaming in and no queries running, none of the meters that drive CCU cost are active, and you’re only billed for storage on whatever’s already materialized. Charges resume as soon as either data starts streaming in or queries start running again.
The Real-Time Context Engine has two meters on the Confluent Cloud Console Billing page: a Storage meter and a Compute meter. Both live under the Kafka cluster resource, metered every minute and rolled up into hourly increments on your bill.