RTK

RTK is an open-source CLI proxy with 84,000+ stars on GitHub that reduces LLM token consumption by 60–90% on common development commands.

Overview

AI coding agents burn a surprising share of their context — and budget — on raw command output: thousand-line test logs, verbose build traces and directory listings that the model mostly does not need. RTK sits between your agent and your shell as a filtering proxy. When the agent runs common developer commands, RTK intercepts the output and compresses it intelligently, keeping the signal (errors, failures, summaries) while stripping the noise, cutting token use by 60–90% on typical commands. It is a single Rust binary with zero dependencies, so it drops into any environment without runtime baggage, and it is designed around agentic coding workflows with tools like Claude Code. The project is Apache-2.0 licensed and very active, with repository updates in October 2026.

Key Features

• 60–90% token savings – compresses common dev-command output before it reaches the model.
• Signal-preserving filters – keeps errors and key results while dropping repetitive noise.
• Single Rust binary – zero dependencies, fast startup, easy installation anywhere.
• Agent-first design – built for Claude Code and other agentic coding workflows.
• Transparent proxying – commands run normally; only the returned output is optimized.
• Cost and context relief – lower API bills and more room in the context window for real work.

💳 Pricing

RTK is completely free and open source under the Apache License 2.0. It saves you money rather than costing any — your only spend remains your model provider’s tokens, now fewer of them.

Who Is It For

Developers who use AI coding agents heavily and feel the token burn from verbose command output — especially teams watching API costs or context limits on long sessions.