Voratiq is an open source engineering framework for producing higher-quality code through structured competition between agents.
Multiple agents implement the same spec independently, and each result goes through automated evals. Then, you review the code and merge whichever is best.
Because model performance is task-dependent, and the variance is large enough that a single default leaves quality on the table.
Running several strong agents and selecting the best result is usually a better bet than trying to predict the winner in advance.
Voratiq works best for tasks where you care about code quality: features, bug fixes, refactors, and tests.
Any agent backed by Codex, Claude Code, or Gemini CLI. You configure which models and providers to run in .voratiq/agents.yaml.
Voratiq runs locally and your code, specs, diffs, and eval results stay on your machine.
The only network calls are from the agents talking to their harness providers (Anthropic, OpenAI, Google).
See the docs or tutorial for setup and workflow details.
If you have questions or need help, reach out at support@voratiq.com.