rtk-ai / rtk

项目摘要

RTK is a high-performance CLI proxy that intercepts and compresses shell command outputs before they reach LLM coding agents, reducing token consumption by 60-90%. Implemented as a single Rust binary with zero dependencies, it supports 100+ commands with sub-10ms overhead.

为什么值得关注

With over 83K stars, RTK addresses a critical pain point in agentic coding: the massive token cost of shell command outputs. Its novel proxy approach is language-agnostic, works with any LLM agent, and delivers dramatic cost savings without requiring changes to existing workflows.

优势

• Dramatic token reduction (60-90%) with minimal latency overhead (<10ms)
• Single Rust binary with zero dependencies ensures easy deployment and portability
• Broad command coverage (100+ commands) handles common dev workflows comprehensively
• Language-agnostic proxy works with any LLM agent (Claude Code, Cursor, etc.) without integration changes

局限性

• 857 open issues suggests potential stability gaps or incomplete edge-case handling
• Limited to shell command output optimization; doesn't address other token consumption sources like conversation history or tool definitions

使用场景

• Reducing API costs for AI coding agents running shell commands
• Extending effective context window capacity by compressing verbose command outputs
• Accelerating agent workflows by reducing token processing latency