9router is a free, MIT-licensed proxy that runs on your own machine and keeps a coding session alive when a provider cuts you off. One npm command opens an endpoint at localhost:20128 that speaks the standard OpenAI format, so Claude Code, Codex, Cursor, or Cline can talk to it unchanged. Behind it, requests fan out across the 60-plus providers it lists in a three-tier fallback: your paid subscription first, then cheap options like GLM and MiniMax, then free tiers like Kiro and OpenCode. When a quota runs dry it switches instead of failing, so a run that would stall at 2am keeps going for free.
It also trims tokens on the way through. A built-in feature called RTK, ported from the Rust “Rust Token Killer” project, compresses tool output like git diff and grep before it reaches the model, which 9router says cuts 20 to 40% of a request’s input tokens without changing the answer. A separate caveman mode shortens replies by up to 65%. For a solo developer who cannot absorb $20 to $50 a month per provider, that means fewer rate-limit walls and a smaller bill.
The convenience has two catches. Routing a paid subscription’s quota through another tool sits in a terms-of-service gray zone, and 9router’s optional bridge that intercepts GitHub Copilot, Antigravity, and Kiro traffic pushes that further; the project says to mind each tool’s policy. A local proxy that holds your keys and sees every prompt is a trust decision, and although the repo is MIT-licensed, its core app ships as a private package you cannot fully audit.
Read More: where 9router fits among six token-cutting tools.
Sources:
- 9router
- 9router source and README (decolua/9router)
- RTK, the Rust Token Killer that 9router ports for compression
- CLIProxyAPI, the Go router that inspired the JavaScript port
Disclaimer: For information only. Accuracy or completeness not guaranteed. Illegal use prohibited. Not professional advice or solicitation. Read more: /terms-of-service
Reuse
Citation
@misc{kabui2026,
author = {{Kabui, Charles}},
title = {9router: {A} {Free} {Local} {Proxy} {That} {Routes} {Around}
{AI} {Rate} {Limits} and {Cuts} {Token} {Costs}},
date = {2026-07-03},
url = {https://toknow.ai/posts/9router-proxy-ai-rate-limits-token-compression/},
langid = {en-GB}
}