Spotify Cut Claude Code Token Use 90% With Shunt

Spotify reports about 90% lower Claude token use on delegated bulk-read work, not on the whole bill.

The shunt plugin uses PreToolUse hooks to block full reads of files over 350 lines and route them to a cheaper model.

Delegation adds 10 to 30 seconds per call and fails on editing and subtle debugging.

You can copy the pattern with a Haiku subagent and your own Read hook.

Read the full post

Read the full post