UncenAI — 1-Click Free T4 GPU Engine for VS Code Copilot Agent-Mode
UncenAI turns free cloud T4 GPU compute (Google Colab, Lightning.ai Studio, and Modal Serverless) into a unified, uncensored, abliterated local coding engine with full agent tool calling (find-replace, multi-find-replace), thinking modes, and 1-click status bar controls directly inside VS Code and Copilot Chat.
Enjoy ~130+ Free T4 GPU Hours per Month with zero local VRAM required.
⚡ The 3-Provider Free T4 GPU Stack
| Provider |
Free Monthly Quota |
Setup & Auth |
Auto-Stop |
Status Icon |
| Google Colab Free |
~3–4 hours / day (~90–120h/mo) |
colab.exe (No credit card) |
20 minutes |
🔥 |
| Lightning.ai Studio |
22 hours / month (15 free credits) |
lightning login (No credit card) |
20 minutes |
⚡ |
| Modal Serverless |
$30 / month free credit (~50h/mo) |
modal setup |
3 minutes |
☁️ |
All models run quantized GGUF on 15 GB cloud T4 VRAM with full native agent mode tools:
- Qwen 2.5 Coder 14B Instruct Abliterated (24k context) — Premier coding reasoning.
- Josiefied Qwen3 14B v3 Abliterated (24k context, Thinking) — Deep multi-turn agent logic.
- Qwen 3.5 9B Abliterated Fast 32k (32k context, Thinking) — Ultra-fast long-context editor.
- Gemma 4 12B Heretic Abliterated (24k context, Thinking) — Uncensored general reasoning.
- Qwen 3.6 35B-A3B MoE Abliterated (12k context, Thinking) — Massive MoE capacity.
- DeepSeek R1 Qwen 14B Abliterated v2 (24k context, Thinking) — Step-by-step reasoning specialist.
🚀 3-Step Quick Start
Install the free cloud CLI tools using uv or pip:
# Using uv (recommended)
uv tool install google-colab-cli
uv tool install lightning-sdk
uv tool install modal
# Or using pip
pip install google-colab-cli lightning-sdk modal
2. Authenticate Your Accounts
Open the VS Code command palette (Ctrl+Shift+P / Cmd+Shift+P) and run:
UncenAI: Login to Cloud GPU Providers
(or run py src/local_controller.py auth --backend colab in your terminal)
3. Register Models in Copilot Chat
Open the command palette and run:
UncenAI: Register Models in Copilot Chat (Agent Mode)
Then click Reload Window when prompted.
🎮 Status Bar Controls & Agent Mode Usage
Look at the bottom-left status bar in VS Code:
⊘ UncenAI [Colab T4]: OFF: Click to open the quickpick menu and boot the GPU.
🔥 UncenAI [Colab T4]: Qwen 2.5 Coder 14B (12m/180m): GPU online! Hover to view quota breakdown across Colab, Lightning.ai, and Modal.
Using in Copilot Chat
- Open GitHub Copilot Chat (
Ctrl+Alt+I / Cmd+Alt+I).
- Switch to Agent Mode (
@agent).
- Select any
UncenAI: ... model in the model picker.
- Prompt Copilot to write code, edit workspace files, or debug complex issues.
🛡️ Quota Protection Guarantee
- Watchdog Auto-Stop: Remote runtimes auto-terminate after 20 minutes of idle time (3 minutes for Modal scale-to-zero).
- Hard Daily / Monthly Caps: Proactively prevents exhausting free quotas.
- Safe Offline Mocking: If Copilot sends a request while offline, a polite notification appears with a 1-click boot prompt.
📄 License
MIT License. Built with ❤️ for the open-source and local AI community.