steipete/oracle

What it solves

Oracle is a CLI tool designed to simplify the process of providing large amounts of context (files and directories) to AI models. It eliminates the manual effort of copying and pasting files into a chat interface by bundling prompts and local files into a single request, allowing users to interact with various LLMs via API or browser automation.

How it works

Oracle operates through two primary engines:

• API Engine: Uses official APIs from OpenAI, Google (Gemini), and Anthropic (Claude) to send prompts and bundled files.
• Browser Engine: Automates a Chrome browser to interact with ChatGPT or Gemini directly, allowing users to use high-end models (like GPT-5.5 Pro) without needing an API key.

It supports multi-model runs where a single prompt is sent to multiple AI models simultaneously for comparison. It also manages sessions, allowing users to follow up on previous conversations using session IDs or response IDs.

Who it’s for

• Developers who need to feed entire codebases or specific file sets into an LLM for review, debugging, or architecture planning.
• Power users who want to compare responses from different AI models (GPT, Gemini, Claude) in one run.
• Users without API keys who prefer to use the browser-based automation to access premium AI models.

Highlights

• Multi-Engine Support: Switch between API and experimental browser automation.
• Multi-Model Advisory Panels: Query multiple models at once and aggregate the results.
• Context Bundling: Use globs and excludes to precisely attach local files and directories.
• Handoffs and Lineage: Continue previous conversations via --followup and track session history in a tree format.
• MCP Integration: Works as a Model Context Protocol (MCP) server for tools like Claude Code and Cursor.
• Manual Fallback: Render and copy bundles for manual pasting when automation is blocked.