Built with Clavesa
Four production properties compute something real with clavesa: two traffic dashboards, a registration funnel, and the deal scores attached to every listing on a used-EV browser. Each one below names its workspace and its pipeline nodes, so the claim can be checked rather than taken.
01 In production
None of these has a data team. Each is one engineer deciding that a number was worth having, and getting it without standing up a warehouse to hold it.
The traffic dashboard on this site is a clavesa pipeline reading this site's own CloudFront logs. There is no analytics vendor behind it and no backend. The raw client IP is dropped inside the pipeline; only a two-letter country code survives to the output.
The deal scores and lifecycle tags attached to every used-EV listing are computed by clavesa rather than by the app. Three SQL transforms run over the accumulated daily scrape history and export the tag file the frontend loads. This is the interesting case on the page: a pipeline producing a feature people use, not a chart the owner looks at.
The mushroom-spot map at oma.tienoo.com and its showcase site share one pipeline, the largest of the four. Alongside the daily rollups it builds retention cohorts and a registration funnel, then writes a separate JSON document for each surface that needs one.
A recipe collection that went live in August 2026 with its analytics already running. Retrofitting measurement is the usual outcome because standing up a pipeline is normally a project of its own. Here it was five nodes copied from an existing workspace and pointed at a new log bucket.
02 Where these came from
Clavesa is a solo project and these are the sites its author runs, so treat this as a record of what the tool does under real load rather than as independent endorsement. The reason it is worth publishing is that each entry is falsifiable: the workspaces are in the repos, the node names are the real ones, and every output listed is a file the live site serves.
Two more sites share the tracker script but do not run a clavesa pipeline. radio.tienoo.com and sceggle aggregate their logs with a plain Node script, which is the right call at their size and the reason they are named here instead of listed above. The tracker is a separate MIT file and works without any of this.
Every site above started from data it was already collecting and throwing away. The pipeline is what turned it into something queryable.