Using OpenRouter Free Models with Pi Coding Agent

I wanted to run Pi without paying per token, so I pointed it at OpenRouter’s free tier. Here’s how I wired it up.

The Setup

Edit your Pi settings file:

nano ~/.pi/agent/settings.json

Add these two lines:

{
  "defaultProvider": "openrouter",
  "defaultModel": "openrouter/free",
  ...
}

That’s it. The openrouter/free router automatically picks the best available free model each time you chat.

What “Free” Actually Means

• No token charges — ever
• Rate limits: 50 requests/day without credits, 1,000/day if you add $10+ in credits
• Models: Llama 3.3 70B, Qwen3 Coder, GPT-OSS 120B, Nemotron 3 Super

The free router picks whichever model is cheapest and least congested at request time.

Why I Like This

Benefit	Details
Zero cost	Never pay per token
Automatic routing	No model picking needed
Easy toggle	Remove defaultModel to go back to Ollama

Gotchas

• API key required — OpenRouter still needs a key even for free models
• Credit balance — You need some credit balance to activate your key, even for free models
• Rate limits — Hit the daily cap? Either wait or add credits for higher limits

Quick Troubleshooting

“Insufficient credits” error?
Your key is valid but your account never bought credits. Go to https://openrouter.ai/settings/credits and add $10 to unlock the higher free-tier limits.

Want to switch back to Ollama?
Remove defaultModel from settings.json or set it to an Ollama model ID.

Need more throughput?
Add $10 in credits at openrouter.ai/settings/credits. That bumps you from 50→1,000 free requests/day.

The CLI Way

If you’re scripting or want to override:

pi --provider openrouter --model openrouter/free

Or pick a specific free model:

pi --provider openrouter --model qwen/qwen3-coder:free

That’s It

Point Pi at openrouter/free, add a key, done. No token costs, automatic model selection, and you can still swap providers anytime.