⚡ Infinite Claude Code

Run Claude Code CLI & Desktop 100% Free forever with real-time SSE streaming, multi-provider auto-failover, and unlimited autonomous subagents.

  ██████╗ ██╗      █████╗ ██╗   ██╗██████╗ ███████╗     ██████╗ ██████╗ ██████╗ ███████╗
 ██╔════╝ ██║     ██╔══██╗██║   ██║██╔══██╗██╔════╝    ██╔════╝██╔═══██╗██╔══██╗██╔════╝
 ██║      ██║     ███████║██║   ██║██║  ██║█████╗      ██║     ██║   ██║██║  ██║█████╗  
 ██║      ██║     ██╔══██║██║   ██║██║  ██║██╔══╝      ██║     ██║   ██║██║  ██║██╔══╝  
 ╚██████╗ ███████╗██║  ██║╚██████╔╝██████╔╝███████╗    ╚██████╗╚██████╔╝██████╔╝███████╗
  ╚═════╝ ╚══════╝╚═╝  ╚═╝ ╚═════╝ ╚═════╝ ╚══════╝     ╚═════╝ ╚═════╝ ╚═════╝ ╚══════╝
                                 [ INFINITE EDITION ]

🌟 Overview

Infinite Claude Code is an ultra-lightweight, zero-latency local proxy engine that turns official Anthropic tools (Claude Code CLI, Claude Desktop, and IDE extensions) into an unlimited, completely free coding powerhouse.

It translates Anthropic Messages API calls (/v1/messages) into OpenAI-compatible format in real-time, routes them through a prioritized pool of the world's best free cloud models (OpenRouter 120B/31B/reasoning, NVIDIA NIM, Groq, Together AI), and provides seamless offline resilience via local Ollama.

🚀 Why Infinite Claude Code?

Feature	Stock Claude Code	Antigravity IDE	Infinite Claude Code
API Cost	💸 $20 - $200+/mo	Free tier limits	100% Free Forever
Streaming	Standard	Web UI only	Native Real-time SSE (Tokens in ms)
Failover	❌ None (hard crash)	❌ None	⚡ Instant Multi-Provider Auto-Failover
Offline Support	❌ Cloud only	❌ Cloud only	🛡️ Local Ollama Fallback
Autonomous Subagents	Limited by credits	Limited	♾️ Unlimited Infinite Concurrency
System Footprint	N/A	Heavy Electron	< 30MB RAM / 0% Idle CPU

✨ 5 Killer Innovations (What Sets Us Apart)

Unlike conventional proxies that rely on heavy Python environments and naive sequential fallbacks, Infinite Claude Code introduces 5 game-changing capabilities:

1. 🐝 Swarm Concurrency Dispatcher

When Claude Code launches multiple subagents in parallel (o abres múltiples sesiones en background con claude --bg), typical routers bottleneck and trigger 429 Rate Limits on a single model. Our dispatcher distributes concurrent subagent calls in parallel across distinct providers (e.g. Subagent 1 ➡️ Nemotron 120B, Subagent 2 ➡️ North Mini Code, Subagent 3 ➡️ Gemma 4). Multiply team productivity by 4x with zero throttle.

2. 🧠 Semantic Task Routing

Rather than blindly querying the first model on a list, our proxy-level classifier inspects the prompt and tools to route work to the ideal engine:

Architecture / Deep Reasoning ➡️ Nemotron 120B / Kimi K3 (Thinking models)
Code Generation & File Edits ➡️ North Mini Code / Qwen Coder (AST specialists)
Terminal Commands & Quick Queries ➡️ Nemotron Lightning (Sub-300ms latency)

3. 🔧 Self-Healing Tool Repair

Free and open-weight models occasionally emit malformed JSON in tool call parameters (unescaped quotes, unclosed brackets). Stock Claude Code aborts with tool execution errors. Our built-in AST sanitizer intercepts and auto-repairs malformed JSON on the fly, raising tool calling reliability to 99.9%.

4. 📊 Cyber HUD & Real-Time "Money Saved" Tracker

Access http://localhost:20129/hud in any browser to inspect our cyberpunk live dashboard:

💰 Live Dollars Saved: Real-time counter of money saved vs official Claude 3.7 Sonnet API prices.
⚡ Swarm Load Monitor: Real-time gauge of active concurrent subagents.
🚀 Live Telemetry Stream: See which provider is answering, tokens generated, and exact latency in milliseconds.

5. 🪶 Zero-Bloat Footprint (Native Node.js)

Competitors require 500MB+ Python virtualenvs (uv, fastapi, C++ wheels). Infinite Claude Code is written in 100% dependency-free Node.js, starts in 10ms, and consumes under 30MB of RAM.

🏗️ Architecture

flowchart TD
    subgraph Client["💻 Coding Environment"]
        CC["Claude Code CLI / Desktop / VS Code"]
    end

    subgraph Router["⚡ Infinite Claude Router (127.0.0.1:20129)"]
        SSE["Anthropic SSE Real-Time Translator"]
        POOL["Dynamic Priority Pool & Auto-Failover"]
        TOOLS["OpenAI Function Calling <-> Anthropic Tool Use"]
    end

    subgraph Cloud["🌐 Cloud Free Providers"]
        OR1["OpenRouter: Nemotron 120B Super (Free)"]
        OR2["OpenRouter: Gemma 4 31B (Free)"]
        OR3["OpenRouter: Nemotron 30B Reasoning (Free)"]
        OR4["OpenRouter: North Mini Code (Free)"]
        NIM["NVIDIA NIM: Kimi K3 Reasoning"]
        GROQ["Groq / Together / Cerebras (Optional)"]
    end

    subgraph Offline["🏠 Local Fallback"]
        OLLAMA["Ollama: Qwen3 / DeepSeek Local"]
    end

    CC -->|POST /v1/messages| SSE
    SSE --> POOL
    TOOLS <--> POOL
    POOL -->|1st Priority| OR1
    POOL -->|Failover #1| OR2
    POOL -->|Failover #2| OR3
    POOL -->|Failover #3| OR4
    POOL -->|Failover #4| NIM
    POOL -->|Failover #5| GROQ
    POOL -->|Offline / Emergency| OLLAMA

🤖 Model & Provider Catalog (49+ Supported AI Providers)

The router supports 49+ model providers grouped into 4 intelligent tiers. If a provider is slow or reaches rate limits, the router fails over to the next in milliseconds:

🌟 Tier 1: 100% Free Cloud Pool (Pre-configured, No API Key needed)

OpenRouter Nemotron 120B Super: nvidia/nemotron-3-super-120b-a12b:free
OpenRouter Gemma 4 31B: google/gemma-4-31b-it:free
OpenRouter Nemotron 30B Reasoning: nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free
OpenRouter North Mini Code: cohere/north-mini-code:free
OpenRouter Nemotron 3.5 Lightning: nvidia/nemotron-3.5-lightning:free
OpenRouter Ultra 550B: nvidia/nemotron-3-ultra-550b-a55b:free
OpenRouter Gemma 4 26B: google/gemma-4-26b-a4b-it:free
OpenRouter Laguna S 2.1: poolside/laguna-s-2.1:free
OpenRouter Apodex 1.1 Mini: apodex/apodex-1.1-mini:free
OpenRouter Liquid LFM 2.5: liquid/lfm-2.5-2.6b:free

⚡ Tier 2: Core Quad-Routers & High-Reasoning Cloud

NVIDIA NIM Cloud: moonshotai/kimi-k3 (Deep reasoning & 16k context)
NVIDIA NIM Foundation: nvidia/nemotron-3-super-120b-a12b
OmniRouter: Meta-router con balanceo y failover automático entre modelos libres (deepseek-chat)
9Router: Despachador de alta velocidad para modelos libres (qwen-2.5-coder)

🔑 Tier 3: Optional Extended Cloud (Press ENTER to Skip during install)

Groq Cloud: LLaMA 3.3 70B & Qwen 2.5 Coder (800+ tokens/sec)
Cerebras Inference: LLaMA 3.3 70B (1,000+ tokens/sec)
SambaNova Cloud: Qwen 2.5 72B & LLaMA 3.3 70B
Together AI: DeepSeek V3 & GLM 5.2
DeepInfra: DeepSeek V3 & V4 Flash
SiliconFlow: Qwen 2.5 Coder 32B
Mistral AI: Codestral Latest
Google Gemini: Gemini 2.5 Flash & 2.5 Pro
xAI: Grok 2 & Grok 4.5
OpenAI: GPT-4o Mini
Anthropic Direct: Claude 3.7 Sonnet / Opus
Plus: Nebius, Chutes AI, Featherless, ZenMux, Agnes AI, W&B, Kilo Code, Cloudflare AI, Alibaba Qwen.

🛡️ Tier 4: Local Offline Resilience

Ollama Local: qwen3:4b / deepseek-r1 (100% offline fallback if internet drops)

⚡ Quick Start & Interactive Installer

The installer provides two modes:

[1] Express Install (Recommended): Sets up everything in 3 seconds using the pre-configured free tier. Zero questions asked.
[2] Custom Install: Lets you paste custom API keys for optional providers. Pressing ENTER skips any provider instantly.

Windows (PowerShell 1-Click)

& ([scriptblock]::Create((irm "https://raw.githubusercontent.com/juanitogit/infinite-claude-code/main/install.ps1")))

Or clone manually:

git clone https://github.com/juanitogit/infinite-claude-code.git
cd infinite-claude-code
powershell -ExecutionPolicy Bypass -File .\install.ps1

macOS / Linux

git clone https://github.com/juanitogit/infinite-claude-code.git
cd infinite-claude-code
chmod +x install.sh && ./install.sh

2. Verify Status

Check that the router is live:

node bin/infinite-claude.js status

Output:

🟢 Infinite Claude Code is ACTIVE and HEALTHY!
📡 URL: http://127.0.0.1:20129
⚡ Streaming: ENABLED (SSE Real-time)
🤖 Active Providers in Pool (9):
   [1] OpenRouter - Nemotron 120B Super (Free)
   [2] NVIDIA NIM - Kimi K3 Reasoning
   [3] OpenRouter - North Mini Code (Free)
   ...

3. Launch Claude Code

Simply open your terminal or IDE and run:

claude

Type anything:

❯ ey bro
¡Hola! ¿En qué puedo ayudarte hoy con tu código? 🚀

Zero billing. Zero tokens burned. Instant responses.

🛠️ Configuration & Custom Keys

You can add your own keys or customize models in src/providers.js or via environment variables:

Variable	Description	Default
`OPENROUTER_API_KEY`	Custom OpenRouter key	Pre-configured Free Tier
`NVIDIA_API_KEY`	Custom NVIDIA NIM key	Pre-configured NIM Key
`OMNIROUTER_API_KEY`	Custom OmniRouter key	Free Auto Endpoint
`ROUTER9_API_KEY`	Custom 9Router key	Free Dispatcher Endpoint
`GROQ_API_KEY`	Optional Groq key for LLaMA 70B	None
`TOGETHER_API_KEY`	Optional Together AI key	None
`OLLAMA_HOST`	Custom Ollama endpoint	`http://localhost:11434`
`PORT`	Local router port	`20129`

🧩 Advanced: Unlimited Autonomous Subagents

Because Infinite Claude Code operates without API cost caps, you can safely unleash Claude Code's multi-agent capabilities and background sessions:

# Ejecutar directamente pidiendo subagentes autónomos
claude "Divide la tarea y usa subagentes concurrentes en paralelo para crear la plataforma completa"

# O despachar múltiples agentes autónomos en segundo plano
claude --bg "Refactor the authentication module and write unit tests"

All subagents will run concurrently, load-balanced across the provider pool without hitting rate limits or draining billing accounts!

👤 Author & Maintainer

Created and actively maintained with ❤️ by JuaneX.

infinite-claude-code