- What kinds of errors can the decoder explain?
- Stack traces, build errors, type errors, runtime exceptions, Kubernetes pod errors, browser console errors, compiler messages, and most other developer error output. Paste the raw text and the decoder handles the rest.
- Do you store the errors I paste?
- No. Inputs are sent to the model in memory and discarded. Nothing is logged, persisted, or used for training.
- Is there a usage limit?
- Yes. To keep the tool free and abuse-resistant we allow 5 decodes per IP every 15 minutes. The window resets automatically.
- Which AI model powers the decoder?
- Anthropic's Claude. The model streams its answer token-by-token so you see the explanation appear as it is generated.
- Can I use this for production debugging?
- Yes, with the usual caution. Treat the suggested fix as a strong hint, not a guarantee. Always verify the root cause before shipping a change.