9Router — One Gateway to 60+ AI Providers with Smart Fallback

Managing multiple AI subscriptions udah jadi masalah sendiri. Tiap bulan gw ada Claude, OpenAI, Gemini, plus some free-tier accounts. Tiap CLI tool (Claude Code, Codex, Cursor) butuh config berbeda. Too many API keys, too many endpoints.

Enter 9Router — one OpenAI-compatible endpoint di localhost:20128 yang nge-route semua requests ke 60+ AI providers dengan 3-tier smart fallback.

What is 9Router?

9Router itu AI gateway yang jalan locally di mesin lu. Dia bikin satu endpoint (OpenAI-compatible) yang bisa dipake oleh semua tools:

┌─────────────────┐
│  Claude Code    │
│  Codex CLI      │
│  Hermes Agent   │
│  Cursor / Cline │
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│  9Router        │ ← localhost:20128
│  (Smart Router)│
└────────┬────────┘
         │
    ┌────┼────┬────┬────┐
    ▼    ▼    ▼    ▼    ▼
┌────┐┌────┐┌────┐┌────┐┌────┐
│Tier1││Tier2││Tier3││Free││OAuth│
└────┘└────┘└────┘└────┘└────┘

3-Tier Smart Fallback

Ini yang bikin 9Router powerful — automatic fallback pas quota habis:

Tier 1 — Subscription (Priority)

• Claude Code (Anthropic)
• OpenAI Codex
• GitHub Copilot
• Gemini (Google AI Studio)

Tier 2 — Cheap & Fast

• GLM Coding ($0.60/1M tokens)
• MiniMax ($0.20/1M)
• Kimi ($9/month unlimited)

Tier 3 — FREE (Unlimited)

• iFlow
• Qwen (Alibaba)
• OpenCode
• Kiro AI
• OpenRouter free tier

Gimana cara kerjanya?
Lu coding pake Claude Code. Pas quota Claude abis, 9Router otomatis switch ke Tier 2 (GLM/Kimi), terus kalau perlu ke Tier 3 (FREE). Lu nggak pernah berhenti coding.

Built-in Token Savers

Dua fitur yang bikin 9Router hemat token (dan uang):

1. RTK (Rust Token Killer) — INPUT Compression

RTK auto-compress tool results (git diff, grep, ls, find) sebelum dikirim ke LLM.

• Lossless compression — sama informasi, cuma lebih ringkas
• Default ON — Jalan otomatis
• Savings: 20% — 65% tergantung command

# Tanpa RTK: 47K tokens
git diff --staged

# Dengan RTK: 28K tokens (same info)

2. Caveman Mode — OUTPUT Compression

Inject terse-style system prompt biar LLM jawab lebih ringkas.

• 5 intensity levels
• "verbose paragraph" → "telegraphic why use many token when few do trick"

My Setup on VPS

Gw install 9Router di VPS (Ubuntu 22.04) biar jalan 24/7 dan bisa diakses dari mana aja.

Installation

bash
Loading...
Loading syntax highlighting...

Running as System Service

Gw pake PM2 buat jagain 9Router supaya auto-start:

bash
Loading...
Loading syntax highlighting...

Cloudflare Tunnel (Access from Anywhere)

Karena 9Router jalan di localhost:20128, gw pake Cloudflare Tunnel biar bisa diakses dari luar:

bash
Loading...
Loading syntax highlighting...

Sekarang gw bisa pake 9Router dari laptop manapun, termasuk Hermes Agent di macOS.

Connecting Tools

Setiap tool tinggal point ke localhost:20128 (atau tunnel URL):

Claude Code:

bash
Loading...
Loading syntax highlighting...

OpenAI Codex:

bash
Loading...
Loading syntax highlighting...

Hermes Agent (config.yaml):

yaml
Loading...
Loading syntax highlighting...

60+ Providers Supported

9Router supports 10 service kinds — bukan cuma chat/LLM:

Service Kind	Examples
Chat/LLM	Claude, GPT-4, Gemini, Qwen, GLM, Kimi
Embeddings	Voyage, Jina, OpenAI, Cohere
Text-to-Speech	ElevenLabs, Deepgram, Edge TTS
Speech-to-Text	Deepgram, AssemblyAI, Whisper
Image Gen	Fal, Stability, BFL Flux, Recraft
Vision/Image-to-Text	OpenAI, Gemini, Anthropic, Groq
Video Gen	Runway ML, Topaz
Web Search	Tavily, Brave, Serper, Perplexity
Web Fetch	Tavily, Exa, Firecrawl, Jina Reader
OAuth Subscription	Claude Code, Copilot, Cursor, Kiro

Real-time Dashboard

9Router punya visual dashboard di http://localhost:20128 (atau tunnel URL):

• Live token tracking per provider
• Quota reset countdown
• Cost estimation
• Provider health status
• 3-tier fallback logs

Gw bisa liat exact usage dan sisa quota sebelum ke-limit.

Why I Love It

1. Zero downtime coding — Pas quota habis, auto-switch ke provider lain
2. Maximize subscriptions — Claude Code, Copilot, Cursor tetep kepake maksimal
3. FREE tier backup — iFlow, Qwen, OpenCode siap cadangan
4. Token savings — RTK + Caveman hemat 20-65% tokens
5. One endpoint — Nggak perlu gonta-ganti API keys di setiap tool
6. Cloud sync — Tunnel + Cloudflare, akses dari mana aja

What's Next

• Setup Smart Combos — chain multiple providers jadi one virtual provider
• Configure multi-account per provider (round-robin load balancing)
• Integrate proxy pools buat region-restricted providers
• MITM Bridge — intercept Antigravity/Copilot/Kiro IDE traffic

---

Building this setup ngajarin gw gimana manage multiple AI subscriptions tanpa pusing. Sekarang gw cuma perlu inget satu endpoint: localhost:20128. Sisanya diserahin ke 9Router. 😊

Stack: 9Router • PM2 • Cloudflare Tunnel • Claude Code • OpenAI Codex • Hermes Agent • RTK • Caveman Mode

Cost: $0/month (pake free tiers + subscriptions lu) — FREE forever buat 8 providers, atau pake subscription lu untuk 6 providers OAuth.