Claude now displays a context usage warning when you resume an existing conversation that's approaching its limit. The update is live in Claude's web interface and appears when picking up a session where a significant portion of the context window is already filled.
Context window refers to the maximum amount of text Claude can hold active at once - think of it as working memory. Claude supports up to 200,000 tokens per session (roughly 400 pages of text), but long conversations, pasted documents, and code files can fill that budget faster than it sounds. When a session runs out of context space, Claude starts dropping the oldest parts of the conversation to make room for new input. Earlier instructions, uploaded files, and decisions made at the start of a session quietly disappear.
The warning gives you that information before you start working, not after you've already run into problems. If you resume a session and see that 85% of the context is used, you know you're better off starting a fresh session rather than continuing - especially if the early parts of the conversation contained important setup, a detailed brief, or a large document.
This is most useful for people who run long working sessions in Claude: researchers pulling apart large documents, developers pasting in entire codebases, anyone running multi-hour projects through a single thread. The signal is small, but the cost of ignoring it - disjointed responses, forgotten instructions, a model that seems to "forget" decisions made an hour ago - is real.
No announcement accompanied the change. It appears to have shipped quietly.