OPEN SOURCE · INTELLIGENT ROUTING

Never Stop Coding. The Free AI Router & Token Saver.

Save 20–40% tokens with RTK + auto-fallback to free and budget AI models for Claude Code, Cursor, Codex, Copilot, and Cline.

Quick Start (npm install) View on GitHub

$npm install -g @rabbi696/ezrouter

# Initialize local proxy with intelligent token compression

$ezrouter start --port 8080 --rtk-mode aggressive

✔ EzRouter proxy active on http://127.0.0.1:8080/v1

✔ RTK Token Saver active · 3-Tier fallback ready

WHY EZROUTER

Stop Burning Subscription Quotas and Overpaying

Standard development setups waste tokens on repetitive context logs and suffer hard rate-limit interruptions.

WITHOUT EZROUTER

The Fragile AI Loop

✕
Subscription quota expires unused
Paid plans like Claude Pro or Cursor reset monthly with wasted unused limits.
✕
Rate limits stop you mid-coding
Hard 429 throttling errors disrupt debugging sessions right when you are close to a fix.
✕
Tool outputs burn tokens fast
Raw compiler dumps and repeated file traces bloat token context windows by 300%.
✕
Expensive fallback APIs
Switching to direct pay-per-token models burns dollars rapidly on simple routine tasks.

WITH EZROUTER

Thoughtful Resilient Routing

✓
RTK & Headroom Token Saver
Compresses repetitive context and strips redundant tool logs, saving 20–40% tokens.
✓
Smart 3-Tier Auto-Fallback
Fails over smoothly from Subscriptions → Low-cost APIs → Completely Free tiers.
✓
Multi-Account & Auto Token Refresh
Rotates across multiple accounts and keeps OAuth tokens continuously renewed in background.
✓
Drop-in Universal Compatibility
Standard OpenAI-compatible endpoint with instant protocol and format translation.

SYSTEM ARCHITECTURE

How EzRouter Smart Routing Works

Your coding tools connect to one local proxy. EzRouter dynamically selects the optimal model path.

SOURCE IDE / CLI

Cursor · Claude Code · Cline

Local Developer Machine

➔

EZROUTER CORE PROXY

RTK Token Compression & Smart Load Balancer

Quota Detection · Auto-Retry · Protocol Translation

Tier 1 Subscription Quotas

Claude Pro, Codex, Copilot $0 Marginal

Tier 2 Cheap Pay-Per-Token

GLM-4, MiniMax, DeepSeek Ultra-low cost

Tier 3 Free Providers

Kiro AI, OpenCode Free, Vertex 100% Free

CORE CAPABILITIES

Built for Uncompromising Developer Flow

Every mechanism is designed to eliminate context bloating and guarantee zero downtime.

RTK Token Saver

Intelligently deduplicates repeating tool outputs and context, saving 20–40% tokens per conversation.

Headroom Saver

Dynamically summarizes past conversation turns when token count approaches model context limits.

Caveman Mode

Enforces dense, concise reasoning responses, substantially reducing completion token costs.

Ponytail Pipeline

Chains lightweight models for initial planning before routing implementation code to heavy models.

Smart 3-Tier Fallback

Zero-latency auto-failover: seamless transition from subscription quotas to budget and free tiers.

Real-Time Quota Tracking

Tracks rate limits and reset windows in real time to avoid triggering unexpected provider throttling.

Format Translation

Adapts OpenAI, Anthropic, Vertex, and Gemini request payloads on the fly without breaking tool calls.

Multi-Account Support

Pool multiple accounts and subscriptions into one aggregate token quota with round-robin balancing.

Auto Token Refresh

Background OAuth refresh handles token expiry automatically without interrupting active sessions.

Custom Combos

Define custom model combinations and priority routing rules tailored for specific IDE workflows.

Cloud Sync

Securely synchronize configuration profiles and provider credentials across multiple developer laptops.

Usage Analytics

Inspect comprehensive token expenditure, savings percentage, and response latency breakdowns.

BROAD INTEGRATIONS

40+ Supported AI Providers & Models

Connect your existing subscriptions, free developer tiers, or custom API endpoints.

OAuth & Subscriptions

6 Platforms

Claude Code OAuth

Antigravity Sub

Codex OAuth

GitHub Copilot Sub

Cursor Sub

Kimchi Sub

Free & Community Providers

100% Free

Kiro AI FREE

OpenCode Free FREE

Vertex AI FREE

Gemini Free Tier FREE

Direct API Key Providers

40+ Providers

OpenRouter

GLM-4

Kimi

MiniMax

OpenAI

Anthropic

Google Gemini

DeepSeek

Groq

xAI

Mistral AI

Perplexity

Together AI

Fireworks AI

+20 more providers

40+

Integrated Providers

100+

Supported AI Models

20–40%

Avg. Token Savings

MIT

Open Source License

UNIVERSAL API INTERFACE

Standard OpenAI-Compatible Proxy

Point your client to http://localhost:8080/v1/chat/completions and EzRouter handles token compression, load balancing, and fallbacks transparently.

✓ 100% SSE Streaming compatibility

✓ Tool & function calling preserved end-to-end

✓ Header-based routing overrides

request.sh — cURL

# Call through EzRouter with smart token saving
curl http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ez-local" \
  -H "X-EzRouter-Tier: auto" \
  -d '{
    "model": "auto-smart-routing",
    "stream": true,
    "rtk_compression": true,
    "messages": [
      {
        "role": "user",
        "content": "Refactor router auth middleware"
      }
    ]
  }'

ENGINEERED WITH MODERN, HIGH-PERFORMANCE FOUNDATIONS

Node.js

Next.js

React 18

Tailwind CSS

SQLite / Better-SQLite3

TypeScript

GET STARTED IN SECONDS

Start Saving Tokens and Coding Uninterrupted

Free forever, open-source, and runs completely locally on your hardware.

$ npm install -g @rabbi696/ezrouter

View on GitHub npm Package