TLDR builds five layers of code analysis—AST, call graphs, control flow, data flow, and program dependence—to give AI agents exactly the code context they need. It indexes a project locally, then answers queries like function relationships or variable origins by retrieving only relevant code. The daemon runs continuously, and the CLI exposes semantic search and context retrieval commands optimized for LLM consumption. Supports 16 languages and claims 95% token reduction versus raw code dumps.