Context window

The context window is how much the model can hold in mind at once, counted in tokens. It covers everything in the request: the system prompt, your files, the conversation so far, and the tools. Every message adds to it.

When a Claude Code session nears the limit, it compacts the conversation: it summarizes the older turns so the thread keeps going instead of losing them. This happens automatically near the limit, or you can run /compact yourself. Either way, detail is lost in the summary, so state that has to survive is safer written down than left in the chat. Run /context to see how much of the window is in use.