Chat with the best zero-cost AI models right inside VS Code. No subscription, no credit card, no paid plan ever required.
Cascade is a lightweight VS Code sidebar extension that connects to three free AI providers — OpenRouter, Hugging Face, and Groq — and lets you chat, ask questions, write code, and work through multi-step problems without spending a penny.
Your API keys are encrypted in VS Code's built-in Secret Storage and never touch any third-party server. Nothing leaves your machine except the messages you send to your chosen AI provider.
| Feature | Details |
|---|---|
| Free models | Llama 3.3 70B, DeepSeek R1, Gemma 3, Qwen 3, Mistral 7B, and more |
| Three providers | OpenRouter · Hugging Face · Groq — switch any time |
| Progress panel | Multi-step tasks show a live progress tracker with step checkmarks |
| Session history | Persistent chat sessions survive VS Code restarts; search and restore any session |
| File context | Attach the active file, open editors, any workspace file, or your Problems list |
| Code actions | Copy, Apply to File, and Run in Terminal buttons on every code block |
| Secure key storage | Keys stored in VS Code Secret Storage — never in settings.json or git |
| Full settings panel | Temperature, max tokens, top-p, context depth, approval mode, file & terminal access |
- Download
NolvusMadeIt.cascade-1.0.0.vsixfrom the Releases page. - In VS Code open the Command Palette (
Ctrl+Shift+P) and run Extensions: Install from VSIX… - Select the downloaded file and reload when prompted.
Search Cascade in the Extensions sidebar or visit the VS Code Marketplace page.
- Click the Cascade icon in the VS Code Activity Bar.
- Click ⚙ Settings → API Keys.
- Paste your free API key for any provider (see links below).
- Click Save — the model list loads automatically.
- Start chatting!
| Provider | Sign-up link | Key format |
|---|---|---|
| OpenRouter | openrouter.ai/keys | sk-or-… |
| Hugging Face | huggingface.co/settings/tokens | hf_… |
| Groq | console.groq.com/keys | gsk_… |
Access to dozens of free-tier models from a single key. Cascade filters to $0/token models by default. Toggle Free models only off in Settings → Models to see all available models.
Highlights: Llama 3.3 70B · DeepSeek R1 · DeepSeek Chat V3 · Gemma 3 27B · Qwen 3 30B · Phi-4 Reasoning
Direct access to the Hugging Face Inference API. Best for open-weight research models.
Highlights: Qwen 2.5 Coder 32B · Llama 3.3 70B · DeepSeek R1 · Gemma 3 27B
Blazing-fast inference on Groq's LPU hardware. Lowest latency of the three.
Highlights: Llama 3.3 70B Versatile · DeepSeek R1 Distill · Gemma 2 9B · Mixtral 8x7B
Open ⚙ Settings from the Cascade sidebar header.
| Section | Key options |
|---|---|
| API Keys | Add/update/clear keys per provider |
| Models | Free-only toggle, fallback provider |
| Sampling | Temperature, max tokens, top-p, chat history depth |
| Workspace | File access (off/read/read+write), file scope, terminal access, git operations, approval mode |
| Profile | Your name (Cascade addresses you personally), custom system prompt |
| Privacy | Chat history retention (unlimited/30d/7d/off), clear all history |
See CONTRIB.md for full setup instructions, project structure, and PR guidelines.
See ABOUT.md for the motivation behind Cascade, the roadmap, and credits.
MIT — see LICENSE for details.
Built with ❤ by NolvusMadeIt