ClaudeGate

Use Claude Code CLI, OpenAI Codex CLI, and OpenCode with ANY AI Model.
A fast, lightweight local gateway that connects your favorite coding agents to DeepSeek, Ollama, OpenAI, Gemini, Groq, OpenRouter, and more — 100% free, local, or cloud.

---

What is ClaudeGate?

CLI coding assistants like Claude Code CLI and OpenAI Codex CLI can read your codebase, write features, run terminal commands, and fix bugs autonomously.

Normally, these tools are locked into expensive proprietary subscriptions.

ClaudeGate is a local bridge for your machine. It sits quietly in the background on port 8082. When your coding agent sends a request, ClaudeGate translates it in real time into standard OpenAI format, sends it to any AI provider of your choice, and streams the response right back.

┌─────────────────────────────────┐
│       Your Coding Agent         │
│  • Claude Code CLI              │
│  • OpenAI Codex CLI             │
│  • OpenCode / Aider             │
└────────────────┬────────────────┘
                 │ HTTP (localhost:8082)
                 ▼
┌─────────────────────────────────┐       ┌─────────────────────────────────────┐
│           ClaudeGate            │       │         Any AI Backend              │
│   Fast Local Gateway Proxy      │ ────► │  • 100% Free / Local: Ollama, vLLM  │
│  • Real-time Tool Streaming     │ ◄──── │  • Ultra-Fast: DeepSeek, Groq       │
│  • Automatic Failover           │       │  • Frontier: OpenAI, Gemini, Claude │
│  • Live Web Console Dashboard   │       │  • Multi-Model Tier Routing         │
└─────────────────────────────────┘       └─────────────────────────────────────┘

You get all the superpowers of elite coding assistants, with total freedom over which AI model powers them.

---

Why Use ClaudeGate?

Without ClaudeGate	With ClaudeGate
Locked to expensive official API subscriptions	Use free tiers, cheap providers, or your existing API keys
All your code is sent to closed cloud servers	Run 100% offline & private on your computer with Ollama
Session crashes if an API rate-limit hits	Automatic Failover: switches to a backup model seamlessly
Works with only one specific CLI tool	Works with Claude Code, Codex CLI, OpenCode, and Aider
Risk of leaking API keys or secrets	Secret Redaction: automatically scrubs passwords & tokens
No visibility into what the proxy does	Live Dashboard: view latency, tokens, and logs in real time

---

See It In Action

ClaudeGate translating tool calls, bash commands, and streaming responses live:

1. ClaudeGate Running in Terminal 1

2. Coding Agent Working in Terminal 2

---

Quick Start (3 Easy Steps)

Step 1: Install ClaudeGate

# Clone the repository
git clone https://github.com/Santosh-Prasad-Verma/ClaudeGate.git
cd ClaudeGate

# Create and activate virtual environment
python3 -m venv .venv
source .venv/bin/activate

# Install dependencies
pip install -r requirements.txt

---

Step 2: Choose Your AI Provider

Pick your favorite provider in one command:

# 100% Free & Local with Ollama (no API key needed):
python start_proxy.py --preset ollama

# DeepSeek Official (Fast & smart coding):
python start_proxy.py --preset deepseek

# Groq (Ultra-fast inference):
python start_proxy.py --preset groq

# Google Gemini:
python start_proxy.py --preset gemini

# OpenRouter (100+ models):
python start_proxy.py --preset openrouter

Or run the interactive setup wizard: python start_proxy.py --setup
(If your provider requires an API key, edit .env and paste your key).

---

Step 3: Connect Your Agent & Start Coding

Option A: Claude Code CLI

Add to ~/.claude/settings.json:

{
  "env": {
    "ANTHROPIC_BASE_URL": "http://127.0.0.1:8082",
    "ANTHROPIC_API_KEY": "sk-claudegate-local"
  }
}

Or export in your terminal: export ANTHROPIC_BASE_URL="http://127.0.0.1:8082"

Option B: OpenAI Codex CLI

export OPENAI_BASE_URL="http://127.0.0.1:8082/v1"
export OPENAI_API_KEY="sk-claudegate-local"

Option C: OpenCode / Aider

export OPENAI_BASE_URL="http://127.0.0.1:8082/v1"
export OPENAI_API_KEY="sk-claudegate-local"

Now start ClaudeGate in Terminal 1:

python start_proxy.py

Open Terminal 2 in your coding project, launch your agent (claude or codex), and start coding!

---

Live Web Console Dashboard

ClaudeGate includes a built-in real-time monitoring console at http://127.0.0.1:8082/dashboard.

• Clean, Minimalist Interface: Zero emojis, crisp SVG icons, and zero distracting animations.
• Flight Recorder: Live log of every request, model routed, duration, tokens, and tools invoked.
• Real-Time Telemetry: Tracks requests, success rate, upstream latency, and token cache savings.
• Ping Test Button: Instant one-click connection and latency test against your active AI provider.

---

System & Model Doctor (--doctor)

Not sure if your setup is working? Run the diagnostic doctor:

python start_proxy.py --doctor

It performs 4 automated health checks:

1. Environment Check: Audits .env configuration, keys, and listening port.
2. Local Engine Discovery: Detects Ollama (:11434), LM Studio (:1234), and vLLM (:8000).
3. Upstream Ping: Tests connection and measures round-trip latency to your provider.
4. Tool Calling Test: Validates that your configured model can actually invoke functions and tools.

---

Supported Providers & Presets

ClaudeGate includes 28+ ready-to-use presets in the presets/ folder:

Category	Provider	Preset Name	Best For
100% Offline & Private	Ollama	ollama	Zero data leaves your computer. Completely free.
	LM Studio	lmstudio	Easy desktop GUI for local models.
	vLLM	vllm	High-speed self-hosted GPU servers.
High Performance Cloud	DeepSeek	deepseek	Outstanding coding & reasoning at low cost.
	Groq	groq	Blazing-fast inference speeds.
	Google Gemini	gemini	Huge context window & fast speeds.
	OpenRouter	openrouter	Access to 100+ models with one API key.
	OpenAI	openai	GPT-4o, o1, and latest OpenAI models.
	Mistral AI	mistral	Mistral Large and Codestral models.
	Alibaba Qwen	qwen	Top-tier open coding models.
	Together / Fireworks	together / fireworks	High-throughput cloud hosting.
Enterprise	Azure OpenAI	azure	Corporate Azure cloud deployments.
	Amazon Q	kiro	Amazon Q developer bridge.

Switch models anytime with python start_proxy.py --preset <name> without restarting your coding agent!

---

Smart Features Under the Hood

• Real-Time Tool Streaming: Zero latency on tool calls — arguments stream directly to your CLI as they are generated.
• Automatic Failover: If your main provider hits rate limits (429) or errors (503), ClaudeGate instantly retries with your backup provider.
• Extended Thinking & Reasoning: Supports DeepSeek R1 and Groq reasoning blocks, while automatically stripping <think> tags so follow-up turns stay clean.
• Anti-Timeout Heartbeats: Emits 15-second keep-alive pings during long reasoning pauses so your terminal connection never drops with ECONNRESET.
• Secret Redaction: Protects AWS keys, GitHub tokens, and private SSH keys from being accidentally sent in prompts.
• Context Overflow Guard: Automatically compacts ancient tool outputs when conversations get very long, preventing Context Length Exceeded errors.

---

Command Cheat Sheet

Command	What it does
python start_proxy.py	Starts the ClaudeGate proxy server
python start_proxy.py --preset <name>	Loads a provider preset (e.g. deepseek, ollama, groq)
python start_proxy.py --doctor	Runs system, local engine, and model capability diagnostics
python start_proxy.py --test	Tests if your upstream provider is connected and responding
python start_proxy.py --setup	Interactive configuration wizard
python start_proxy.py --version	Shows current version
curl http://127.0.0.1:8082/dashboard	Opens the web console dashboard
curl http://127.0.0.1:8082/health	Checks proxy health status

---

Running with Docker (Optional)

# Start container in background
docker compose up -d --build

# View live logs
docker compose logs -f

# Stop container
docker compose down

---

FAQ

Do I need a paid Anthropic subscription or API key?

No! You do not need an Anthropic subscription. ClaudeGate lets you power Claude Code CLI completely with other providers (like DeepSeek, Gemini, OpenAI, or 100% free local models like Ollama).

Can my agent still edit files, run bash commands, and run tests?

Yes! ClaudeGate translates all tool calls and bash commands seamlessly between your CLI agent and the backend model. File edits, terminal commands, and searches work just like native Claude.

How do I run 100% free and offline?

Install Ollama, pull a model (e.g., ollama pull qwen2.5-coder:32b or deepseek-r1), and run:

python start_proxy.py --preset ollama
python start_proxy.py

Zero API keys, zero dollars spent, and zero data leaves your machine.

How do I switch to a different model?

Run python start_proxy.py --preset <name> (e.g. deepseek, gemini, groq), then restart python start_proxy.py. Your agent in Terminal 2 will automatically use the new model.

---

License

This project is licensed under the MIT License — open and free for both personal and commercial use.

Built for developers who love CLI coding agents and demand complete freedom of choice.

_{If ClaudeGate helps your coding workflow, consider starring the repo on GitHub!}