Cut your input tokens before they reach the model.

The Distillery builds local token-optimization tools for AI coding workflows. Our proxy sits between your coding agent and the model API, deduplicates repeated input context, distils large tool outputs, and reports the overall input reduction across your sessions.

The Distillery is a local proxy for developers using AI coding tools.

It is designed for teams and individuals who want lower model bills without changing how they work:

• No prompt rewriting
• No workflow changes
• Local request optimization
• Input-token reduction focused on real coding sessions
• Usage reporting for sessions, savings, and team visibility

The primary package is thedistillery.

npm install -g thedistillery
thedistillery auth login
thedistillery start

Then use your AI coding tool normally.

The Distillery works with Claude Code automatically and with other tools that can point at a custom Anthropic or OpenAI-compatible base URL.

Current focus:

• Claude Code
• Cline
• Goose
• Aider
• Zed
• Codex CLI
• OpenAI-compatible agent tooling

• Website: https://thedistillery.dev
• App: https://app.thedistillery.dev
• Install guide: https://thedistillery.dev/install
• Client guides: https://thedistillery.dev/clients
• Agent setup guides: https://thedistillery.dev/for-agents

Optimization runs locally on your machine. The hosted service receives usage metadata for account, billing, dashboard, alerts, and attribution features. It does not receive your code or prompts for optimization.

Pinned repositories below are the best place to start.