Baraklabs

Run Claude Code For Free

OpenRouter + Ollama setup 🚀

Claude Code is one of the most powerful AI coding tools right now.

It can read your codebase, edit files, and even execute commands.

But there’s one problem…

👉 It’s expensive and locked to Anthropic models.

Developers want flexibility — not restrictions.

The Problem

  • Only works with Anthropic models out of the box
  • High API costs for long coding sessions
  • No flexibility to use better or cheaper models

The Solution

Use OpenRouter and Ollama to unlock multiple models and run Claude Code without paying Anthropic.

Step 1: Install Claude Code

Install the CLI using the method appropriate for your operating system. If you have used Claude Code before, run /logout first.

Mac / Linux

curl -fsSL https://claude.ai/install.sh | bash

Windows (PowerShell)

irm https://claude.ai/install.ps1 | iex

Alternatively, use npm:

npm install -g @anthropic-ai/claude-code
  • Run claude to initialize.
  • Log in via the default Anthropic flow to authorize the CLI.
  • You can close the "Add Credit" screen; once authorized, the terminal will show "Login successful."

Step 2: Configure OpenRouter

Create an API Key in OpenRouter > Activity > API Keys. Then, override the default settings using one of the methods below:

Method A: Shell Profile (Permanent)

Add these to your ~/.zshrc or ~/.bashrc:

export OPENROUTER_API_KEY="your-key-here"
export ANTHROPIC_BASE_URL="https://openrouter.ai/api"
export ANTHROPIC_AUTH_TOKEN="$OPENROUTER_API_KEY"
export ANTHROPIC_API_KEY="" # Must be explicitly empty
export ANTHROPIC_MODEL="openrouter/free"
export CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS=true

Method B: Project Level

Edit .claude/settings.local.json:

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://openrouter.ai/api",
    "ANTHROPIC_AUTH_TOKEN": "<your-openrouter-api-key>",
    "ANTHROPIC_API_KEY": "",
    "ANTHROPIC_MODEL":"openrouter/free",
    "CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS": "true"
  }
}

Step 3: Pick a Model

OpenRouter allows you to use various models. If you get a 402 error, check your balance or switch to a free model.

Model NameBest For
GPT OSS 120B (Free)Basic testing / No credits
NVIDIA NemotronCoding and agentic tasks
Minimax 2.5UI components & Web Generation

Tip: Use /model inside Claude Code to switch models on the fly.

Step 4: Verify & Build

Restart your terminal and run claude. If it asks for permissions, choose "Yes, allow all" to avoid repeated prompts.

Verification Prompt:

"Hi, how are you doing today?"

App Building Prompt:

Create a weather application showing rainy weather for Guwahati, stormy for Kohima, and sunny for Delhi. Include animations and a clean UI with navigation.

Alternative: Ollama Cloud

For faster performance and ideation, you can use the Ollama launch flow. This doesn't require high-end local hardware as models run in the cloud.

  • Ensure Ollama is installed locally.
  • Use the --model flag to specify cloud models.
ollama launch claude --model minimax-m2.7

You don’t need premium models for everything.

Most developers just need something that works well.

And now — you can do it for free.

Recommended Blogs