Compose a PRView SourceOpen Issue

How KubeStellar Console Is Built: An AI-Maintained Codebase

March 2026

KubeStellar Console is of the first open source projects where the majority of code is written, reviewed, and maintained by AI agents — with human oversight at every step.

This isn’t about replacing developers. It’s about what becomes possible when AI handles the volume and humans focus on direction.

The Numbers

In the last two weeks alone, over 90 PRs were merged into the console repository. Each was:

  • Generated by a coding agent (Claude, Copilot, Gemini, or Codex)
  • Reviewed by an AI review agent for code quality
  • Validated by automated tests and security scanners
  • Approved by a human maintainer

The result: bug fixes ship in hours instead of weeks. Security issues found by scanners get patched the same day. New features go from idea to merged PR in a single session.

How It Works

Bug-to-Squash

When you report a bug through the console’s feedback system (the bug icon in the top bar, or navigate to /issue):

  1. A GitHub issue is created automatically
  2. A coding agent analyzes the issue and creates a PR
  3. A review agent checks the PR for quality
  4. Automated tests validate the fix
  5. A human maintainer approves and merges
  6. You get notified when the fix is live

Feature-to-Fulfillment

Feature requests follow the same pipeline. Describe what you want, and the system figures out how to build it.

Continuous Quality

Beyond user-reported issues, automated QA runs hourly:

  • Auto-QA scans the codebase for bugs, type errors, and inconsistencies
  • Security scanners check for vulnerabilities
  • Nilaway catches nil pointer dereferences in Go code
  • Consistency tests verify the card registry, translations, and API contracts

When these scans find issues, they automatically create GitHub issues, which trigger the same AI fix pipeline.

Contributing

The best way to contribute is through test PRs. A failing Playwright test or a new API contract test tells the AI agents exactly what to build. Tests define the expected behavior — the agents write the implementation.

See CONTRIBUTING.md for details.