Skip to content
The Distillery

Your Claude Code bill grows every session

Context accumulates. Each turn costs more than the last. The Distillery stops the compounding.

Why every session costs more

Every Claude Code message re-sends your conversation history. A 20-message session sends 20× the tokens of message one. The Distillery sits between your tools and Anthropic and distils repeated context before the request leaves your machine.

Measured reduction

30-60% cost reduction depending on session patterns. The 20% figure is the conservative deterministic benchmark floor on standard fixtures.

20%
standard sessions (benchmark floor)
38%
heavy agentic sessions
74%
monorepo exploration

Run it yourself: see the benchmark →

How it works

1
Install
npm install -g thedistillery && thedistillery auth login
2
Start
thedistillery start
3
Claude Code is auto-configured

thedistillery start writes ANTHROPIC_BASE_URL into ~/.claude/settings.json via hooks. No shell profile edit needed — just restart Claude Code after starting the proxy.

That's it. Your tools keep working exactly as before.