Developer open-sources ClaudeCut to cut Claude API bills by 40-70%

On October 5, 2026, a developer posting on DEV Community announced that ClaudeCut is now open source. ClaudeCut is described as a lightweight local reverse proxy that sits in front of Anthropic's API and is claimed to cut Claude API and Claude Code bills by 40% to 70% without any drop in code quality.

In the announcement, the author wrote that anyone building with Claude Code, Cursor, or autonomous agentic workflows over the past few months has likely felt the sting of runaway Anthropic API bills, noting that a few days of multi-turn coding sessions can easily run up hundreds of dollars in token usage.

The mechanism is straightforward: ClaudeCut automatically inserts the cache_control breakpoints that Prompt Caching requires into outgoing requests, so the very long context that gets resent on every turn is billed at cached rates instead of full price.

The 40% to 70% figure comes from the author's own account, and the report explicitly states that no third-party verification is available. The source does not specify which models or versions are covered, nor how the proxy handles requests that already carry caching breakpoints. The project ships as a local proxy rather than a hosted service, and the post frames the cost problem as one that shows up specifically in long, multi-turn coding sessions.