Claude Code with OpenRouter

- 3 minutes read - 504 words

I have been using GitHub Copilot and Copilot CLI for about one year. During that time, I also experimented with several developer tools and AI-assisted coding technologies. Recently I became interested in exploring Claude Code and related tools such as Claude Cowork.

However, I did not want to immediately commit to a subscription before understanding whether Claude Code actually fits my workflow. So I spent some time experimenting with different ways to run Claude Code without paying for a subscription first.

This post summarizes what I tried and what worked best for me.

Option 1: Running Claude Code with Local Models

The first approach I experimented with was running local models and connecting them to Claude Code. Tools such as:

  • Ollama

  • vLLM

  • LM Studio

make it possible to run local LLMs and expose them via APIs.

To support this experiment, I purchased a small AI mini PC:

GMKtec NUCBOX EVO-X2 AI (Ryzen AI MAX+395)

My idea was to run models such as:

  • qwen3

  • qwen3.5

  • qwen3-coder

and connect Claude Code to them.

Example commands using Ollama:

ollama launch claude --model qwen3-coder
ollama launch codex --model qwen3-coder

Unfortunately, the experience was not very good.

Problems I encountered:

  • Inference was quite slow, especially for coding tasks.

  • Models sometimes produced errors or unstable responses.

  • The overall workflow was not smooth enough for daily development.

For experimentation this setup was interesting, but for serious usage it was not practical.

If someone wants to run coding models locally long-term, better options might include:

  • Renting GPU instances in the cloud

  • Buying a high-end GPU workstation

  • Using specialized inference setups

For my mini PC, the performance was simply not strong enough.

So I decided to move on to another approach.

Option 2: Using OpenRouter with Claude Code

The second approach turned out to be much more practical.

Instead of running local models, Claude Code can be configured to use OpenRouter. OpenRouter acts as a unified API layer that allows access to many different models.

Advantages of using OpenRouter:

  • No subscription required to try different models

  • Ability to experiment with multiple providers

  • Pay-as-you-go pricing

  • Easy integration with Claude-compatible APIs

To configure Claude Code with OpenRouter, I exported the following environment variables:

export OPENROUTER_API_KEY="sk-or-v1-"
export ANTHROPIC_BASE_URL="https://openrouter.ai/api"
export ANTHROPIC_AUTH_TOKEN="$OPENROUTER_API_KEY"
export ANTHROPIC_API_KEY=""
export ANTHROPIC_MODEL="deepseek/deepseek-v3.2"

With this configuration, Claude Code can interact with models provided by OpenRouter, such as:

  • DeepSeek

  • Claude models

  • Other frontier models supported by OpenRouter

This setup allowed me to quickly test multiple models and evaluate how well they perform for coding tasks.

Final Thoughts

If you want to try Claude Code without committing to a subscription immediately, there are a few possible approaches.

Local models:

  • Good for experimentation

  • Requires strong hardware

  • Often slower for coding workflows

OpenRouter:

  • Much easier to set up

  • Supports many models

  • Flexible and cost-effective for testing

For my workflow, OpenRouter is currently the most practical way to explore Claude Code before deciding whether a full subscription is worth it.

I will continue experimenting with different models and workflows to see how well Claude Code integrates into my development environment.

comments powered by Disqus