Published: 2026-06-12

Run Claude Code Free with Nex-N2: Plug a Free API Into the Claude Code Harness

Chapters / key moments (click to jump — plays here on the page)

Julian Goldie demonstrates running the Claude Code harness for free by swapping in a free API "brain" — the new Nex-N2 (Next N2 Pro) model, available free on OpenRouter with a 262K-token context window. He uses the open-source free-claude-code project (the same way Ollama plugs local models into Claude Code) and adds a voice layer to speak prompts and watch apps build. The same free model also drops into Hermes agent, Kilo Code and Pi.

Source video

"Free Claude Code + N2 is INSANE!" by Julian Goldie SEOWatch on YouTube →

Key Takeaways

  • The core idea: a free "brain" in the Claude Code harness. The open-source free-claude-code project lets you plug any API into Claude Code — the same pattern as wiring Ollama into Claude Code — so you keep the harness but pay nothing for the model.
  • Nex-N2 is the free model used. In the video it's described as agentic, multimodal (text + images), built for coding, tool use, deep research and long-horizon workflows, with a 262K context window and free access on OpenRouter. (Note: Goldie calls N2 "Claude 3.5 architecture," while creator Fahd Mirza describes it as a ~400B Qwen 3.5 MoE — treat the architecture claim as unconfirmed.)
  • Voice + skills layer. Goldie wires a voice control on top so a single spoken sentence builds an app, and notes you can teach free-claude-code skills just like Claude Code (e.g. a video-generation skill).
  • Not just Claude Code. The same free N2 model can be used directly via OpenRouter, or plugged into Hermes agent, Kilo Code, or Pi — since June 8 it's reportedly been used most with Claude Code and Hermes.
  • Honest caveats (from the creator). "Free" isn't unlimited — it gets rate/token-limited and may stop mid-build; and it is not as capable as paying for a frontier model directly. Good for testing ideas and lightweight builds, not a frontier replacement.
Before you install anything

The setup relies on a third-party open-source project and a free hosted model. Review the project's source and the model's terms before connecting API keys, and prefer the official open-source repo over any paid "done-for-you" bundle. See our security guide and responsible-AI notes.

Commands & Code Mentioned

# Conceptual flow shown in the video (no exact CLI was displayed):
# 1. Get the free Nex-N2 (Next N2 Pro) model on OpenRouter — 262K context, free tier
# 2. Use the open-source `free-claude-code` project to plug that API into the Claude Code harness
#    (same pattern as connecting Ollama to Claude Code)
# 3. Optionally add the voice layer, then speak a prompt: e.g. "create a to-do list app"
# 4. Or use the same model directly in Hermes agent / Kilo Code / Pi via the OpenRouter endpoint

The video is a screen walkthrough rather than a terminal tutorial; the steps above summarise the wiring it demonstrates.

Weekly Digest — In Your Inbox

Get the week's top AI agent news, updates, and guides — every Friday.