Repo Digest.

Everything you need to know about your codebase. One call chain on Qwen 3.8 or GPT Luna, three documents, and the bill at the end.

Context budget. The repo is downloaded and its files are ranked (README, docs, config, then source, tests last) and packed into one prompt until the budget is hit. Larger budgets read more of the codebase per document and cost more. Each of the three documents re-sends that prompt, so a 500K run is three ~500K-token calls. Qwen3.8 Flash has a 1M window at $0.16 / $0.47 per M; Max has 262K at $2 / $6, so its budgets stop at 240K. GPT-6 Luna and GPT-5.6 Luna bill by prompt size, with a higher tier past 272K tokens.