Nothing to hand? Load the
— a retired date-suffixed model, a hardcoded sk-ant- key, a
budget_tokens thinking config, the old output_format parameter,
a JSON-forcing prefill, tiktoken doing the cost math and no caching anywhere
— or the ,
where the correct verdict is api-current and the useful output is what to do next.
Paste the integration — the prescan is free
No upload, no AI: the prescan reads your code in the browser and lists what it mechanically found — every model ID, SDK import, call surface, server tool and beta flag — then the flags: retired or date-suffixed model IDs, budget_tokens, the deprecated output_format parameter, sampling params on models that removed them, assistant prefill turns, requests with a stable system prompt but no cache_control, tiktoken counting Claude tokens, outdated web_search variants, fast mode missing its beta flag or its supported model, huge max_tokens on non-streaming calls, hardcoded sk-ant- keys, one broad exception handler and mixed provider SDKs. Each group explains why it matters. This part costs nothing and happens while you type.
The AI reviews it — this is the metered part
A senior integration engineer's pass, grounded in the current Claude API surface: an API posture with the single most important change named, the inventory with each construct's role, prioritized findings across model drift, deprecated usage, correctness, caching, cost, reliability and hygiene — each with the problem, what it actually costs in dollars or failed requests, the fix and a corrected fragment in your language — and then the ordered plan to get there. Every prescan flag is confirmed or explicitly set aside. Findings may only cite code that appears in what you pasted. Pricing is honest: a worst-case amount is reserved before the run and only what the run actually uses is charged — the meter next to the button shows both.
Fix, export, re-run
The plan as a commented shell script, every corrected fragment in one paste-ready block, the findings as GitHub PR review comments with ```suggestion fences, a tickable action checklist for the ticket, the findings table as CSV, and Markdown or JSON export of the whole review. Then start fixing: the prescan re-runs as you edit and the strip above the run button counts the flags you have cleared, the ones still open and any you have just introduced — in the browser, for free, before you pay for a second review. Reviews are saved to your SkillSafe account when you are signed in, so they follow you to another machine; restore puts the code back in the form too.
Derived from the @anthropics/claude-api skill. Not affiliated with Anthropic; reviews run on the platform's configured model.