Weekly ClickHouse Cloud spend digest posted to Slack

Every Monday at 8am, build a plain-language ClickHouse Cloud cost digest for our data team and post it to Slack. Make this an agent workflow, because it has to interpret the numbers, decide what counts as a notable spike, and write a readable summary rather than dump raw figures.

First, pull the organization usage cost report twice with ClickHouse Cloud's Get Organization Usage Costs operation: once for the last 7 days (the seven days ending yesterday) and once for the prior 7 days (the seven days before that). Pass from_date and to_date in YYYY-MM-DD form for each window. ClickHouse Cloud bills on two dimensions, compute (memory-hours) and storage (compressed data), so keep those two components separate as you read the report.

Then call List Services to get the current ClickHouse Cloud services, so you can map each service id in the cost report to its human-readable name and label the spend per service.

Interpret the two reports and work out: total spend this week versus last week including the percentage change; the per-service breakdown; and the split between compute and storage. Call out any service whose cost rose more than 25 percent week over week as a notable spike.

Write a short, plain-language digest that a non-technical teammate can skim: lead with total spend this week versus last week and the direction of the trend, then the per-service breakdown, then the compute-versus-storage split, and finally any spike callouts. Keep it a readable narrative, not a raw data dump.

Post the finished digest to our #data-infra Slack channel using the Slack Bot Send a Message operation. Format it with Slack mrkdwn so it is easy to read at a glance.

Handle the edge cases gracefully: if a service appears in one week but not the other, note it as newly added or removed rather than reporting a misleading percentage change, and if the cost report comes back empty for a window, say so plainly instead of guessing.