cache-tax
Keeps the prompt cache warm during breaks and stops a cold send once — with its price.
Source and capability notes checked · By Karan Bansal · Updated Oct 23, 2026
What cache-tax looks like in use
No recording yet for cache-tax. The repository README has the details.
Before you install
A mod is not sandboxed. It runs inside Claude Code with your permissions. This one:
- can read Claude settings
- can read Every prompt you send
- can spend Your usage allowance
Check it yourself before running it: git clone https://github.com/karanb192/cache-tax && claude plugin validate ./ · MIT licensed
Install
claude plugin marketplace add karanb192/claude-code-mods
claude plugin install cache-tax@claude-code-mods
Then run /keepwarm inside the session.
Needs Claude Code 2.1.287 or later. After installing, run
/reload-plugins if the session was already open.
Full install guide
The problem it names
Prompt caching is the reason a long session is affordable. The cache expires after roughly an hour of inactivity, and the next message you send re-writes the entire context at the full input price.
You do not see this happen. You see a session that suddenly cost more.
What it does
Two things, both narrow:
Keeps the cache warm while you are away. A cheap periodic touch so that coming back after lunch does not mean re-reading 300k tokens from scratch.
Warns once before a cold send, with the price. Not a block, not a setting to hunt down — one message with the number, so you decide whether to compact first.
Why it is not just “always keep warm”
Because always-on warming costs tokens too. There is a break-even: short gaps are not worth paying to keep, long ones are. The tool is useful precisely because it puts the number in front of you instead of guessing for you.
When to skip it
On short sessions this does nothing. If you open Claude Code, finish a task and close it inside twenty minutes, you never hit a cold send and never see the point.
Also worth knowing: if you already run a dashboard that shows cache state, this duplicates the warning without adding the break-even math.
Similar mods
context-inspector
Finds repeated reads, similar outputs and compaction-related waste.
cache-keeper
Combines cache timing, cold-send warnings, chat board and handoff.
agent-usage-guard
Holds a burst of subagents, a huge prompt or a retry loop before it eats your plan window.
idle-compact
Compacts once after fifty idle minutes, before cache expiry.
Popular mods
agent-profiler
Breaks a session into model wait, tool and subagent time.
Agent Time Machine
Records tool calls and project snapshots on a branching timeline.
loop-detective
Collects evidence when the same Bash failure keeps repeating.
pin-memo
Keeps personal notes and todos above the prompt across resumes.