Context Window Optimizer
Plan what goes into a model’s context window before you ship. Paste each part of a request, or just enter its size: system prompt, tool definitions, retrieved documents, chat history, and the question. See whether the reply still fits, what each request costs, and what to trim first.
How to use it
- For each section, paste real text or type an approximate token count. Use counts for things like long histories you don’t want to paste.
- Set the window size and price for the model you use. Check your provider’s current numbers.
- Read the advice. It points at the biggest section and the cheapest fix.
What’s in each request
Model
Repeated prefixes like system prompts can often be cached. This assumes cached tokens cost about 10% of the normal price; use 0 if unsure.
Budget
- Window used
- 0%
- Per request
- $0
- Per month
- $0