Already Have Claude Code in Your Terminal? Here's How to Run It Free.
This guide assumes you've already got Claude Code installed and running in your terminal. Three steps. No Anthropic API costs. Ever.
GitHub Repo → github.com/Alishahryar1/free-claude-codeNot sure what Claude Code is or how to get it set up? Head over to theaisurfer.com — it's where people who are serious about leveraging AI to build better lives and businesses actually hang out. Come learn to ride the wave.
How It Works
A developer on GitHub named Alishahryar1 built a free proxy tool called free-claude-code. It sits between Claude Code and the internet, intercepting API calls and rerouting them to free AI models — while Claude Code thinks it's still talking to Anthropic. The models it routes to (DeepSeek, Kimi, GLM) run at roughly 80–90% of Claude's performance. The cost: zero dollars.
Over 25,000 developers are already running this setup.
Step 1 — Install the Proxy
Open a new terminal window (keep your existing Claude Code terminal separate for now).
macOS / Linux:curl -fsSL "https://github.com/Alishahryar1/free-claude-code/blob/main/scripts/install.sh?raw=1" | sh
irm "https://github.com/Alishahryar1/free-claude-code/blob/main/scripts/install.ps1?raw=1" | iex
What this does — Pulls and installs the proxy tool on your machine. Takes about 30 seconds depending on your connection.
Step 2 — Pick Your Free AI Provider and Add Your Key
You have three options. Pick one.
Option A — OpenRouter (recommended for most people)- Go to openrouter.ai and create a free account
- Navigate to API Keys and generate a new key
- Free models available include DeepSeek, Kimi, and GLM — no credit card required for free-tier models
- Go to build.nvidia.com and sign up for a free account
- Generate an API key at build.nvidia.com/settings/api-keys
- Gives you free access to hosted open-source models including DeepSeek
- Requires a decent GPU (8GB VRAM minimum recommended)
- Install Ollama if you haven't already
- Pull a model locally — DeepSeek works well for coding tasks
Once you have your key, open the local Admin UI to configure your provider:
- Open a browser and go to
http://127.0.0.1:8082/admin - Paste your API key into the relevant provider field and save
Step 3 — Start the Proxy and Launch Claude Code
In your terminal, start the proxy server:
fcc-server
Validation checkpoint — You should see a confirmation message in the terminal that the proxy is running. If it hangs or throws an error, go back to http://127.0.0.1:8082/admin and confirm your API key is saved correctly, then retry.
Then launch Claude Code through the proxy:
fcc-claude
This replaces your normal Claude Code launch command. The fcc-claude command automatically points Claude Code at the local proxy — no manual environment variable needed.
What to Expect
- Performance — Expect 80–90% of Claude's output quality on most coding tasks. Complex multi-file reasoning is where you'll notice the gap most. Routine tasks, refactoring, and boilerplate generation hold up extremely well.
- Speed — Varies by provider and model load. OpenRouter free-tier models can occasionally be slower during peak hours.
- Cost — Zero. Free-tier API keys on OpenRouter and NVIDIA NIM have usage limits but are genuinely free with no billing required.
Don't have Claude Code yet? This guide won't make much sense without it. Jump over to theaisurfer.com — a community of people learning to use AI tools like this to build real things. We break it all down from scratch.