Vercel launches CI built for the flood of coding-agent pull requests

Vercel has launched Vercel CI in beta, pitching it as continuous integration designed for repositories where coding agents create more pull requests and make test wait times a growing bottleneck.

The service reads a project’s existing turbo.json configuration and uses Turborepo’s dependency graph to run only the tasks affected by a code change. Unaffected work can be restored from Remote Cache, while each selected task runs on compute sized for that job. Vercel says this made CI two to six times faster in its internal repositories, though that figure is the company’s own benchmark.

Developers and agents can inspect runs, logs and failures through the Vercel dashboard, the vc ci command-line interface or Vercel MCP. There is no separate workflow YAML for projects already using Turborepo.

Vercel says the system has completed nearly 8 million tasks across its own repositories and the private beta. The public beta is available on all plans. The broader significance is that a major deployment platform is treating AI agents as first-class code contributors and rebuilding the testing loop around their higher volume of changes.