Claude Code Budgeting Tools

Use tokentrim to trim Claude Code output, read file ranges, and spool long output to disk. Install as a plugin or clone the repo.

How it works for a single Claude Code session

  1. Install the plugin or clone the repo.
  2. Run your Claude Code script.
  3. The toolkit trims the tool output and reads only the needed file ranges.
  4. If the output is long, it is spooled to disk instead of staying in memory.
  5. After the run, the replay command shows your savings and token usage.

What you get / what it costs

FAQ

How much can I save? On 135 real conversations, history re‑read was cut by 75%, limits lasted 4.0x longer, and billed cost fell roughly 55‑70%. Shrinking tool output alone saved 6.9%.

Do I need a special OS? Python 3.9 or newer is required. Tested on macOS and Linux.

Can I use it on multiple machines? Yes, Pro lets you use it on up to 3 machines.

Next Step

Visit the Get started page to install or clone tokentrim and begin trimming.

Updated 2026-11-01