Claude Code token-saving tips: 5 settings that keep cache hit rate above 98%
AlchainHust · x · 2026-10-13
AlchainHust shares hard-won Claude Code token-saving settings, with a screenshot showing 5.8B tokens used in 7 days and 97% of the weekly quota burned:
- Use Opus5.5 as the main model but don't fill the 1M context: set auto-compaction between 400k–600k; too short hurts quality, too long is slower, pricier, and actually worse.
- Finish tasks in one continuous burst: the main agent's cache lasts 1 hour; coming back hours later means cache misses and much higher input token costs.
- Use Sonnet 5.5 for subagents: Opus adds little for grunt work, Haiku barely saves anything.
- Manually raise subagent cache TTL to 1 hour: the default is only 5 minutes, generating tons of uncached input tokens.
- Launch as few subagents as possible: a solo main-agent conversation sustains 98%+ cache hit rate; any subagent noticeably increases consumption.
More from coding & agent
- Dev spins up a horde of Claudes to hit a one-week deadline — "it worked out" — rickasaurus · 2026-10-13
- Debug Multi-Agent Workflows: Permissions, Handoffs and Channels Beat Model Blame — Al_Grigor · 2026-10-13
- iOS Dev Mocks the AI-Agent Era: HomeView, agents.md, Skills and App Store Rule 2.3bc — Dimillian · 2026-10-13
- Who decides when AI evaluator and human reviewer disagree on an agent call? — AssociationNew7925 · 2026-10-13
- Dev builds real-time dictation pipeline: AssemblyAI Sync API plus LLM Gateway rewrite — tech__unicorn · 2026-10-13
- 'Spinbusy': Yacine mocks coding agents stuck in pointless fix loops — yacineMTB · 2026-10-13