Claude API returns different output than Claude.ai for the same prompt

What's happening

Claude.ai is a product built on top of the API. It supplies its own system prompt, tool access and sampling defaults, so identical user text can produce different output.

Fix it

1. Match the model version exactly — the app may run a newer snapshot.
2. Add an explicit system prompt describing role, format and constraints.
3. Set temperature and max_tokens deliberately rather than relying on defaults.
4. Reproduce any file or search context the web app had access to.
5. Pin a dated model snapshot in production so behaviour stops drifting.

Still stuck?

If output varies between identical API calls, lower temperature and specify the response format explicitly.