A big Claude Code session left idle doesn't auto-compact, so when you return and send one line the whole swollen transcript re-bills at full input price (the cache went cold while you were away). Turn on idle compaction — and tune its token floor — so the history shrinks in the background before you resume.
Let a Long Session Compact Itself While It's Idle, Not When You Resume
🔒 Pro tip · Intermediate
Unlock this tip — and 142 more
This is one of 143 advanced, fact-checked tactics reserved for Pro. Get the full 165-tip library, a searchable archive, and a new tip every morning. Free for 7 days, then $9/mo.
Prefer to browse? The 22 Beginner tips are free forever.
More in Context Management
🧠Context Management
Images can be 1,000-2,000+ tokens each; removing stale ones cuts that per turn
Drop the Screenshot Once the Model Has Read It
Images, PDFs, and attachments are charged as tokens and re-sent every turn in a multimodal thread. After the model has described or transcribed one, you usually don't need to keep sending the pixels.
🧠Context Management
50-90% on file-heavy prompts
Paste the Function, Not the Whole File
Most coding questions need 20-40 lines, not your 800-line file. Send the relevant slice plus a one-line note about the rest, and your input shrinks dramatically without hurting the answer.
🧠Context Management
Trims re-sent history; often 20-60% fewer input tokens per turn after a topic switch
Start a New Chat When the Topic Changes
Chat apps re-send your whole conversation with every message. When you switch tasks, the old turns become dead weight you keep paying to re-transmit — even with caching discounts.