oracle

Oracle is a CLI tool designed to help developers bundle a prompt and selected files into a single request, which is then processed by another model for review, debugging, refactoring, design checks, or cross-validation. By utilizing either the API or browser engine, Oracle allows for the review process to be more efficient, providing real repository context during the analysis. The outputs are intended to be advisory, meaning users are encouraged to validate the results against the actual codebase and tests for accuracy and completeness. This workflow is especially helpful for tasks such as code review, debugging complex scenarios, or ensuring cross-validation across different models.

Oracle offers a range of useful features to streamline this process. Key options include the ability to specify which files to include or exclude using flexible globbing patterns. The tool also supports both browser and API engines, with the browser engine being the default and offering an extended workflow with GPT‑5.2 Pro. Oracle's 'dry-run' mode helps users preview their requests before execution, allowing them to identify potential issues such as excessive token usage. The system supports sessions, which can be stored for future use, ensuring that work is not lost. Additionally, Oracle supports remote execution for running automation on a signed-in machine, adding flexibility to the review process.

Oracle is aimed at developers who need real-time reviews, debugging, or checks with the context of their actual codebase. It can be particularly useful for teams working on large projects or those who need to perform cross-validation between different models. Whether for software engineering, refactoring tasks, or technical design validation, Oracle offers a reliable method to improve the efficiency of these processes by leveraging AI-based reviews with real repository context.