A second CI vendor has to earn it

Technology

CircleCI sells control over the run. You pick a machine size per job, split a test suite across parallel containers using how long each test actually took last time, and reuse one pipeline definition across every repository in the organization.

That matters once the pipeline has become a product of its own, with its own maintainers and its own failure modes. Until then, the CI already attached to the repository is the honest default, and leaving it there is the cheaper answer.

Monogram logo

The pipeline

When a dedicated CI platform is worth the seam

  1. Test splitting is the feature you pay for

    Timing data from previous runs splits a suite across parallel containers so each one finishes at roughly the same moment, and resource classes let a heavy job take a bigger machine without upgrading everything else. Wait time on a pull request is the number this optimizes.

  2. Config that survives more than one repository

    Orbs package a pipeline step once and version it, so twenty repositories share one deploy job instead of twenty copies drifting apart. That is an organization scale problem, and it is the honest reason to run CI somewhere other than next to the code.

  3. A second vendor is a real cost

    Another account, another set of deploy credentials, another status page to check when a build fails for no visible reason. Our repositories are on GitHub and Actions is already there, which is a high bar to clear. CircleCI has not shown up in a build on this site.