Oracle is a CLI and MCP server that sends developer-selected files and a custom prompt to AI models for code review and analysis.
The tool solves the problem of getting grounded AI feedback on actual project code. It bundles files you specify with a prompt, sends that context to an AI model, and stores the result as a session. You can preview the exact prompt and file contents before sending anything, and the tool supports multiple AI providers through either API keys or browser-based authentication. This approach keeps the AI focused on your specific codebase rather than general knowledge.
Oracle suits developers and coding agents who need a second opinion on code changes, architectural decisions, or debugging problems. It works well for teams using multiple AI providers, since you can run the same context against different models in a single command. The tool distinguishes itself by supporting both API mode (for OpenAI, Anthropic, Gemini, xAI, OpenRouter, and compatible endpoints) and browser mode (for ChatGPT and Gemini through Chrome automation), letting you choose based on your setup and preferences. The render mode lets you inspect and copy the bundle without contacting any model, useful for understanding exactly what context will be sent.
The maintainers respond to issues and pull requests within a day. Work in the issue tracker centers on reproducible issues and high-priority items, suggesting a focus on stability and addressing concrete problems rather than exploratory features.
