Deep Search Mode

AI-powered codebase exploration.

Deep Search answers questions about your codebase. Ask how something works, where something is implemented, or what code touches a specific area — and get back structured results with exact file and line references.

What it does

Deep Search explores your codebase to answer a specific question. It reads files, follows dependencies, traces execution paths, and builds an annotated picture of everything relevant to your query. Results are structured and precise — not a summary, but a map of exactly where the answer lives in your code.

Deep Search is entirely read-only. It never modifies any file.

When to use it

• "How does the authentication flow work in this codebase?"
• "Where is the payment logic implemented?"
• "What files are involved in the data export feature?"
• "Which components depend on this shared utility?"
• Onboarding to a new codebase and understanding how it's structured
• Before a refactoring, to understand the full blast radius of a change

How it works

1. Ask your question — in plain language, describe what you want to understand about the codebase.
2. Exploration phase — the agent searches for relevant files, reads their content, and traces dependencies and call chains.
3. Results compiled — findings are collected and organized by relevance.
4. Annotated output — the answer is returned with specific file paths, line numbers, and explanatory annotations for each relevant piece of code.

Key features

• Read-only — Deep Search never modifies any files; it only reads and analyzes
• Annotated results — every finding includes file path, line number, and context explaining why it's relevant
• Dependency tracing — follows imports, call chains, and data flows to build a complete picture
• Structured output — results are organized by relevance, not just listed
• Broad scope — searches the full codebase, not just the files you've recently touched

Tips