🤖 OpenCode + OpenRouter — Free AI Coding Agent on Windows

Run a 120B parameter AI coding agent from your terminal — for free. No GPU, no expensive subscription, no local hardware requirements.

---

📺 Watch the Full Video

I built a complete production landing page — 3 files, animations, pricing tiers, dark UI — in under 2 minutes. Zero cost. This repo is the complete setup guide.

---

⚡ What Is This?

This repo contains everything you need to set up OpenCode as a fully autonomous AI coding agent connected to OpenRouter's free tier — so you can run a 120B parameter model from any machine without a GPU.

Tool	Role
💻 OpenCode	Agentic coding CLI — reads, writes, runs your code
🔀 OpenRouter	Model gateway — one API key, 100+ models
🟢 NVIDIA Nemotron 3 Super	120B free model — 262K context window
🖥️ VS Code	Your editor — PowerShell terminal

---

🖥️ Requirements

Requirement	Detail
OS	Windows / Mac / Linux
Node.js	v18 or higher
Terminal	PowerShell (Windows) / Bash (Mac/Linux)
GPU	❌ Not required
API Key	Free — openrouter.ai

---

🚀 Installation Guide

Step 1 — Install Node.js

🪟 Windows

Download and install from nodejs.org

Verify:

node --version

Should show v18 or higher.

🍎 macOS

brew install node
node --version

🐧 Linux

curl -fsSL https://deb.nodesource.com/setup_20.x | sudo -E bash -
sudo apt-get install -y nodejs
node --version

---

Step 2 — Install OpenCode

npm install -g opencode-ai

Verify install:

opencode --version

Should show 1.14.24 or higher.

---

Step 3 — Create Project Folder

🪟 Windows (PowerShell)

mkdir opencode-demo
cd opencode-demo

🍎 macOS / 🐧 Linux

mkdir opencode-demo && cd opencode-demo

---

Step 4 — Get Your OpenRouter API Key

1. Go to openrouter.ai
2. Create a free account — no credit card required
3. Go to API Keys → Create
4. Name it opencode-demo
5. Copy your key — starts with sk-or-v1-

---

Step 5 — Set Your API Key

🪟 Windows — Session Only (PowerShell)

$env:OPENROUTER_API_KEY = "sk-or-v1-YOUR-KEY-HERE"

🪟 Windows — Permanent (Recommended)

[Environment]::SetEnvironmentVariable("OPENROUTER_API_KEY", "sk-or-v1-YOUR-KEY-HERE", "User")

Then close and reopen your terminal.

🍎 macOS / 🐧 Linux — Session Only

export OPENROUTER_API_KEY="sk-or-v1-YOUR-KEY-HERE"

🍎 macOS / 🐧 Linux — Permanent

echo 'export OPENROUTER_API_KEY="sk-or-v1-YOUR-KEY-HERE"' >> ~/.zshrc
source ~/.zshrc

For bash:

echo 'export OPENROUTER_API_KEY="sk-or-v1-YOUR-KEY-HERE"' >> ~/.bashrc
source ~/.bashrc

Verify:

echo $env:OPENROUTER_API_KEY   # Windows
echo $OPENROUTER_API_KEY       # Mac/Linux

---

Step 6 — Explore Available Models

List all models:

opencode models

Filter free models only:

# Windows
opencode models | Select-String ":free"

# Mac/Linux
opencode models | grep ":free"

---

Step 7 — Configure OpenCode

OpenCode config is stored globally at:

OS	Path
Windows	C:\Users\<username>\.config\opencode\opencode.json
Mac/Linux	~/.config/opencode/opencode.json

Open config:

# Windows
code $env:USERPROFILE\.config\opencode\opencode.json

# Mac/Linux
code ~/.config/opencode/opencode.json

Paste this config:

{
  "$schema": "https://opencode.ai/config.json",
  "model": "openrouter/nvidia/nemotron-3-super-120b-a12b:free",
  "provider": {
    "openrouter": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "OpenRouter",
      "options": {
        "baseURL": "https://openrouter.ai/api/v1",
        "apiKey": "{env:OPENROUTER_API_KEY}"
      },
      "models": {
        "nvidia-nemotron-3-super": {
          "name": "nvidia/nemotron-3-super-120b-a12b:free",
          "limit": {
            "context": 262144,
            "output": 32768
          }
        }
      }
    }
  }
}

⚠️ The apiKey field reads from your environment variable — your actual key never sits in plain text.

---

Step 8 — Test Your Setup

opencode run "Who are you and what model are you running on?"

You should see a response confirming OpenCode is running on NVIDIA Nemotron 3 Super via OpenRouter. ✅

---

🖥️ Using the TUI (Interactive Mode)

Launch the full terminal interface:

opencode

Key Commands Inside TUI

Command	Action
/models	Switch model mid-session
/theme	Change UI theme
/skills	Browse available skills
/sessions	Manage multiple sessions
/undo	Undo last agent action
/redo	Redo undone action
/thinking	Toggle model reasoning visibility
Ctrl+P	Open command palette
Esc x2	Interrupt current task
Tab	Toggle Plan / Build mode

---

🎯 Demo — Test Your Setup

Quick Test — Python Script

opencode run "Create a Python script called hello.py that prints Hello from OpenCode"

Full Use Case — Landing Page

Run this in your project folder:

opencode run "@
ACT AS: Senior Frontend Developer & UI/UX Designer.

TASK: Create a professional 3-file landing page for 'PitchZero Intelligence' — 
an AI-powered platform tracking venture capital funding and startup investment signals.

BRANDING:
- Theme: Deep Obsidian black (#0a0a0a)
- Accent: Electric Ultraviolet (#8b5cf6)
- Tone: High-tech, minimalist, data-driven

SECTIONS:
1. Hero with headline and CTA
2. Intelligence Stack: Live Funding Tracks, Agentic Logic Analysis, Sector Velocity
3. PitchZero Feed: mock blog/newsletter layout
4. Pricing: Free, Pro, Enterprise tiers

FILES TO GENERATE:
- index.html (Semantic HTML5)
- style.css (Flexbox/Grid, neon-glow class)
- app.js (scroll animations, newsletter signup)

OUTPUT: Write all 3 files completely.
@"

Open in browser:

# Windows
start .\index.html

# Mac/Linux
open index.html

---

🔧 How It Works

You type a prompt in OpenCode
        ↓
OpenCode sends HTTPS request to OpenRouter API
        ↓
OpenRouter routes to NVIDIA Nemotron 3 Super (free tier)
        ↓
Model thinks, reasons, generates response
        ↓
OpenCode acts — creates files, runs commands, edits code
        ↓
Loop continues until task is complete

No GPU on your machine. No local model download. Everything runs in the cloud via OpenRouter.

---

📊 OpenRouter Free Tier — What to Know

Item	Detail
Cost	$0.00 per request on free models
Rate limit	~20 requests/minute
Daily cap	Token budget per model
Data policy	Free tier prompts may be used to improve models
Fix if you hit limits	Add $5 credits — goes a very long way

💡 Monitor your usage at openrouter.ai/activity

---

💡 Tips

• Hit rate limit mid-task? Wait for daily reset or add $5 credits
• Wrong model ID error? Use the full ID: openrouter/nvidia/nemotron-3-super-120b-a12b:free
• API key not found? Close and reopen terminal after setting permanent env variable
• Want a different free model? Run opencode models | grep ":free" and update opencode.json
• Sensitive data? Don't use free tier endpoints — prompts may be logged by providers

---

📁 Project Structure

opencode-demo/
├── opencode.json          # Optional per-project config
├── README.md              # This file
└── outputs/
    ├── index.html         # Generated landing page
    ├── style.css          # Generated styles
    └── app.js             # Generated JavaScript

---

🤝 Contributing

Found an issue or want to add more model configs? Pull requests welcome.

---

📄 License

MIT License — free to use, modify and share.

---

🔗 Links

---

Zero Cost. No GPU. Any Machine.
Built with OpenCode + OpenRouter + NVIDIA Nemotron 3 Super