OrcaFlow

AI Router by TechTide AI

One endpoint for all your AI providers. Smart fallback, auto-routing, zero downtime.

Features • Quick Start • Providers • Configuration • Contributing

---

Credits & Acknowledgments

OrcaFlow is a customized fork of 9Router by decolua and contributors.

We give full credit to the 9Router project for the core routing engine, provider integrations, and architecture that makes this possible.

Original Project	github.com/decolua/9router
9Router Website	9router.com
Original License	MIT (preserved in this fork)
Contributors	View all contributors

---

What is OrcaFlow?

OrcaFlow is a local AI proxy/router that sits between your coding tools and AI providers. It gives you:

• One endpoint (http://localhost:20128/v1) for all AI providers
• Smart 3-tier fallback: Subscription → Cheap → Free, automatically
• Format translation: OpenAI ↔ Claude ↔ Gemini, seamlessly
• Multi-account support: Round-robin between accounts per provider
• Usage analytics: Track tokens, costs, and trends locally
• Zero telemetry: Everything runs locally, no data leaves your machine

Your CLI Tool (Claude Code, Codex, Cursor, Cline...)
        |
        v
   OrcaFlow (localhost:20128)
   - Format translation
   - Quota tracking
   - Auto token refresh
        |
        +---> Subscription tier (Claude Code, Codex, Gemini CLI)
        +---> Cheap tier (GLM, MiniMax, Kimi)
        +---> Free tier (iFlow, Qwen, Kiro)

---

What's Different in OrcaFlow?

OrcaFlow builds on 9Router with the following customizations by TechTide AI:

Change	Description
Custom branding	OrcaFlow identity throughout the UI
Security hardening	Re-enabled header masking in request logger to prevent token leaks
Cloud sync disabled	No phone-home capability active
Team customizations	Provider configs, UI themes, and workflow optimizations

---

Quick Start

From source (recommended)

git clone https://github.com/Alexi5000/9router_OrcaFlow.git
cd 9router_OrcaFlow
cp .env.example .env
npm install
npm run dev

Dashboard opens at http://localhost:20128

Production mode

npm run build
PORT=20128 npm run start

---

Supported CLI Tools

Works with all major AI coding tools:

Tool	Status
Claude Code	Fully supported
OpenAI Codex	Fully supported
Cursor	Fully supported
Cline / RooCode	Fully supported
Continue	Fully supported
GitHub Copilot	Fully supported
OpenClaw	Fully supported
Gemini CLI	Fully supported

Configuration (same for all tools):

Endpoint: http://localhost:20128/v1
API Key:  [copy from OrcaFlow dashboard]
Model:    cc/claude-opus-4-6  (or any supported model)

---

Supported Providers

OAuth Providers (Free login)

Provider	Models	Quota
Claude Code	Opus, Sonnet, Haiku	5h + weekly reset
OpenAI Codex	GPT-5.2, GPT-5.1	5h + weekly reset
Gemini CLI	Gemini 3 Flash, 2.5 Pro	180K/month FREE
GitHub Copilot	GPT-5, Claude 4.5	Monthly reset
Cursor	Various	Per subscription

Free Providers (Unlimited)

Provider	Models	Cost
iFlow AI	8+ models (Kimi K2, Qwen, GLM, DeepSeek)	$0
Qwen Code	Qwen 3 Coder Plus/Flash	$0
Kiro AI	Claude Sonnet/Haiku	$0

API Key Providers (40+)

OpenRouter, GLM, Kimi, MiniMax, OpenAI, Anthropic, Gemini, DeepSeek, Groq, xAI, Mistral, Perplexity, Together AI, Fireworks, Cerebras, Cohere, NVIDIA, SiliconFlow, and 20+ more.

---

Key Features

Feature	Description
Smart 3-Tier Fallback	Auto-route: Subscription → Cheap → Free
Real-Time Quota Tracking	Live token count + reset countdown
Format Translation	OpenAI ↔ Claude ↔ Gemini seamless
Multi-Account Support	Multiple accounts per provider with round-robin
Auto Token Refresh	OAuth tokens refresh automatically
Custom Combos	Create unlimited model combinations
Usage Analytics	Track tokens, cost, trends locally
MITM Proxy	Intercept CLI tool traffic for transparent routing

---

Configuration

Environment Variables

Variable	Default	Description
PORT	20128	Service port
JWT_SECRET	(auto-generated)	JWT signing secret
INITIAL_PASSWORD	123456	First login password
REQUIRE_API_KEY	false	Enforce Bearer API key on /v1/* routes
ENABLE_REQUEST_LOGS	false	Enable request/response logs
OBSERVABILITY_ENABLED	true	Local usage tracking

See .env.example for all available variables.

---

API Reference

Chat Completions

POST http://localhost:20128/v1/chat/completions
Authorization: Bearer your-api-key
Content-Type: application/json

{
  "model": "cc/claude-opus-4-6",
  "messages": [{"role": "user", "content": "Hello"}],
  "stream": true
}

Compatibility Endpoints

Endpoint	Format
POST /v1/chat/completions	OpenAI format
POST /v1/messages	Anthropic format
POST /v1/responses	OpenAI Responses format
GET /v1/models	List all models

---

Tech Stack

Component	Technology
Runtime	Node.js 20+
Framework	Next.js 16
UI	React 19 + Tailwind CSS 4
Database	LowDB (JSON file-based, local only)
Streaming	Server-Sent Events (SSE)
Auth	OAuth 2.0 (PKCE) + JWT + API Keys

---

Security

OrcaFlow includes security hardening over the base 9Router:

• No telemetry: No analytics SDKs, no tracking pixels, no beacons
• Cloud sync disabled: The cloud sync feature is disabled and commented out
• Tunnel disabled: The Cloudflare tunnel feature is disabled by default
• Header masking: Sensitive headers (authorization, API keys, tokens) are masked in logs
• Local-only data: All usage data, credentials, and logs stay on your machine
• Proper .gitignore: .env, data/, ecosystem.config.*, logs/ are all excluded

See SECURITY.md for our security policy and vulnerability reporting.

---

Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

---

License

MIT License — see LICENSE for details.

This project is a fork of 9Router which is also MIT licensed. The original copyright notice is preserved.

---

Upstream

This project is built on top of 9Router by decolua. All credit for the core routing engine, provider integrations, format translation layer, and architecture goes to the original 9Router team and contributors.

---

_{Built by TechTide AI on top of 9Router}