Connect Claude Code to NVIDIA-hosted models using your own API key. No Anthropic account needed.
Claude-NIM is a lightweight API gateway that bridges Claude Code with NVIDIA-hosted models. It translates requests and responses between the two formats so everything works transparently.
Point Claude Code at your instance and everything works — streaming, tool use, file uploads, all handled automatically.
Use your own NVIDIA NIM key. Requests and data flow through your account with your own quotas and limits.
Responses stream back as you type — no waiting for complete replies. Supports thinking blocks and extended reasoning.
Upload images and documents. They're processed and passed to the model automatically.
Keys are never stored or logged. Each request is authenticated and isolated per user.
Built-in keepalive pings keep your instance warm on serverless platforms.
Switch between any NIM-supported model by changing a single config value.
Deploy Claude-NIM, configure Claude Code, and start coding — it's that simple.
Run the Docker container or deploy to your preferred platform. The service listens on port 8000 out of the box.
Create an API key from the NVIDIA Build dashboard. Your key starts with nvapi-.
Point Claude Code to your deployment URL and set your NIM key as the auth token. Choose the NIM models you want to use.
{
"env": {
"ANTHROPIC_BASE_URL": "https://your-instance.example.com",
"ANTHROPIC_AUTH_TOKEN": "nvapi-...",
"ANTHROPIC_DEFAULT_SONNET_MODEL": "meta/llama-3.1-70b-instruct"
}
}
Model names are passed through as-is to NIM. Use the ANTHROPIC_DEFAULT_*_MODEL env vars to map Claude Code's model tiers to NIM model IDs of your choice.
Once configured, Claude Code works exactly as you'd expect. All features route through your instance transparently.
Open Claude Code and start a session. It connects to your instance automatically — chat, edits, and tool calls all work seamlessly.
Upload images and documents through the Files API. They're processed and sent to the model as appropriate inputs.
Use the models endpoint to see which NIM models are available to your key, with context windows and pricing details.
curl -X POST https://your-instance.example.com/v1/messages \ -H "anthropic-version: 2023-06-01" \ -H "x-api-key: nvapi-..." \ -H "content-type: application/json" \ -d '{ "model": "meta/llama-3.1-70b-instruct", "max_tokens": 256, "messages": [{"role": "user", "content": "Hello!"}] }'