Anthropic Reveals How Claude Code Bills You — 21 Ways to Cut the Cost
Roger_M_Taylor · x · 2026-09-19
Anthropic has published how Claude Code actually bills: nothing is stored server-side between turns, so every turn re-sends the entire thread — turn 40 pays to re-read the 39 before it, and your own message is the smallest part.
Charlie Hills distilled 21 cost-cutting tactics, 12 of which are single commands:
- /clear between tasks, /compact at phase breaks
- /rewind trims recent turns for free
- /autocompact 200k as a safety net on 1M-context models
- /context to inspect what's loaded; /mcp to disable unneeded servers
- @-mention files instead of typing paths (each mention attaches a copy)
- Set /model up front — each model has its own cache
The takeaway: context is the billing driver, so actively managing session history is direct savings.
More from coding & agent
- Docling, the Document Parser for Gen AI, Tops 66K GitHub Stars — docling-project · 2026-09-19
- Codex-X Adds a Visual Management Panel for OpenAI Codex Desktop and CLI — yynxxxxx · 2026-09-19
- 'Letting Agents Rip' on Your Codebase: The Coding Agent Meme Everyone Gets — charles_irl · 2026-09-19
- Claude Code adds AGENTS.md support in 2.1.277, falling back when CLAUDE.md is absent — eyishazyer · 2026-09-19
- Open-source Jev-cu makes Codex computer use faster by passing text, not screenshots — alexcovo_eth · 2026-09-19
- Jev read 384 news stories for $0.19 while Claude Opus 5 managed 4 for $0.77, claims demo — airesearch12 · 2026-09-19