Repository understanding enables AI to analyze entire codebases, comprehending architecture and dependencies. Why it matters: Real coding tasks require understanding how files interact, not just single-file context. Large codebases exceed context windows. Approaches: Indexing: Parse and index all files, retrieve relevant context for queries. Embeddings: Embed code chunks, retrieve semantically similar code. Graph construction: Build dependency graphs, call graphs, inheritance hierarchies. Summarization: Generate summaries per file/directory, hierarchical understanding. Key capabilities: Answer questions about codebase, navigate to relevant code, understand system design, identify dependencies, trace data flow. Tools: Sourcegraph Cody, Cursor codebase chat, GitHub Copilot Workspace, Continue, custom RAG systems. Technical challenges: Keeping index updated, handling massive repos, choosing retrieval scope, context window limits. Implementation patterns: Hybrid of symbol indexing + semantic search + LLM reasoning. Use cases: Onboarding new developers, impact analysis for changes, architectural understanding, finding similar code. Essential capability for truly intelligent coding assistants.