Anyone doing serious pair programming with Claude Code (Anthropic's terminal coding agent) or OpenCode has inevitably run into the infamous wall:
HTTP 429: Too Many Requests / Quota Exceeded.
When you are deep in a flow state refactoring a complex codebase, having your terminal session freeze because you hit an hourly or daily token limit completely breaks your momentum.
To solve this once and for all, I built OMNICLAUDE — a lightweight, zero-dependency AI Command Center & local proxy designed to keep Claude Code running continuously without interruption or expensive pay-as-you-go API bills.
The Core Problem with Single-Key AI Coding
While Google offers an astonishing 1,000,000 free tokens per day on Gemini Flash, a single API key still faces two major bottlenecks:
- Tokens-Per-Minute (TPM) bursts: Complex Claude Code tool-calls can easily exceed the short-term burst ceiling.
- Quota exhaustion: After extensive sessions, that single key is blocked until midnight UTC.
When that happens, your CLI agent crashes and you are forced to wait.
Enter OMNICLAUDE: Architecture & Concept
OMNICLAUDE is built with three fundamental engineering principles:
-
Zero External Dependencies: Pure native Python (
server.py). No heavy Docker containers, no complex npm daemon processes, no bloat. -
Dynamic Multi-Key Hub & Quota Rotator: Connect multiple free Google AI Studio accounts. The engine tracks daily token consumption with live visual bars (
token_counter). The moment Key #1 returns an HTTP 429 rate limit, OMNICLAUDE executes an automatic 0.1-second failover to Key #2 without interrupting Claude Code or dropping your active context. - Modular Backup Parachute: If cloud providers hiccup, your requests automatically cascade down a customizable priority chain — from Gemini to Groq (Llama 3.3 70B), DeepSeek, or straight into 100% offline local models running on your own GPU via Ollama.
Key Features Breakdown
1. ⚡ Live Gemini Quota Tracking & Hot-Swapping
The dashboard displays your real-time consumption against Google's daily quota:
- Instant test buttons to verify key validity directly against Google API endpoints.
- Auto-recovery of cooldown keys once the rate-limit window expires.
- Visual pool overview (e.g.,
Pool tot: 656k / 5.000k).
2. 🛡️ The Modular Parachute Gateway
Configure fallback priority with intuitive arrow controls:
- Cloud Engines: Google Gemini 2.5 Flash, Groq 70B, OpenRouter, DeepSeek V3/R1.
- Local Offline Models: Automatic auto-detection of any GGUF/Ollama model installed on your PC.
3. 🎨 Obsidian Cyber Interface
Seamless one-click switching between Dark Cyber (pure obsidian with neon mint & cyan accents) and Light Pro (clean cyber-lab theme), preserving your preferences locally.
Quick Start (Under 60 Seconds)
Getting started on Windows 10/11 requires zero complicated setup:
bash
# Clone the open source repository
git clone https://github.com/aton-fux/OMNICLAUDE.git
cd OMNICLAUDE
# Launch the orchestrator (launches the local web UI at http://127.0.0.1:3000)
START.bat
Open http://127.0.0.1:3000 in your browser.
Enter your free Gemini API key in the configuration tab.
Click "🚀 APRI CLAUDE CODE" or "⚡ APRI OPENCODE" and start coding without quota fear!
Open Source Community & Pro Edition
OMNICLAUDE is open-source under a protected GPLv3 license for personal use.
📂 GitHub Repository: aton-fux/OMNICLAUDE
⭐ Show your support: If this saves you hours of coding friction, please consider dropping a Star ⭐ on GitHub!
⚡ INFINITY Edition: For developers running heavy automated agent workflows who need unlimited multi-key pooling and continuous hot-swapping, check out the OMNICLAUDE INFINITY Store
.
Happy coding with uninterrupted agentic AI!
Top comments (0)