Configure NRP LLM with OpenCode and Crush

UPDATED 2026-01-31: Added details about namespace requirements for generating tokens.

The National Research Platform (NRP) provides a free Large Language Model (LLM) service for researchers. This service allows you to access powerful models like GLM 4.7, Llama 3, and Qwen 3 via an OpenAI-compatible API.

This tutorial shows how to configure two popular CLI AI tools, Crush and OpenCode, to use the NRP LLM service.

Get your Token

First, you need to generate an API token. Go to https://nrp.ai/llmtoken/ and log in with your institutional credentials to get your free token.

Note: A user has to be a member of a namespace with LLM feature in order to be able to generate tokens. SDSC users are privileged and are added to sdsc-llm namespace by a daemon every half an hour, but others might need to join a namespace first. Faculty members can have their own namespaces and add others with no restrictions. For more information on namespaces, see Getting Started.

In your terminal, export your token to a variable (replace ... with your actual token):

export TOKEN="your_token_here"

Configure OpenCode

OpenCode is a powerful terminal-based coding assistant.

To configure it for NRP, run the following command to create the configuration file:

mkdir -p ~/.config/opencode
cat <<JSON > ~/.config/opencode/opencode.json
{
  "\$schema": "https://opencode.ai/config.json",
  "provider": {
    "nrp": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "NRP Nautilus",
      "options": {
        "baseURL": "https://ellm.nrp-nautilus.io/v1",
        "apiKey": "$TOKEN"
      },
      "models": {
        "glm-4.7": { "name": "GLM 4.7" },
        "gpt-oss": { "name": "GPT-OSS" },
        "llama3-sdsc": { "name": "Llama 3 (SDSC)" },
        "qwen3": { "name": "Qwen 3" },
        "gemma3": { "name": "Gemma 3" }
      }
    }
  }
}
JSON

Verify it works by running:

opencode run -m nrp/gpt-oss "Hello"

Configure Crush

Crush is another excellent CLI tool.

To configure it for NRP, run this command:

mkdir -p ~/.config/crush
cat <<JSON > ~/.config/crush/crush.json
{
  "providers": {
    "nrp": {
      "id": "nrp",
      "name": "NRP Nautilus",
      "type": "openai",
      "base_url": "https://ellm.nrp-nautilus.io/v1",
      "api_key": "$TOKEN",
      "models": [
        { "id": "glm-4.7", "name": "GLM 4.7", "context_window": 256000, "default_max_tokens": 4096 },
        { "id": "glm-v", "name": "GLM V", "context_window": 256000, "default_max_tokens": 4096 },
        { "id": "gpt-oss", "name": "GPT-OSS", "context_window": 128000, "default_max_tokens": 4096 },
        { "id": "llama3-sdsc", "name": "Llama 3 (SDSC)", "context_window": 128000, "default_max_tokens": 4096 },
        { "id": "qwen3", "name": "Qwen 3", "context_window": 128000, "default_max_tokens": 4096 },
        { "id": "gemma3", "name": "Gemma 3", "context_window": 128000, "default_max_tokens": 4096 }
      ]
    }
  },
  "models": {
    "large": { "model": "glm-4.7", "provider": "nrp" },
    "small": { "model": "gpt-oss", "provider": "nrp" }
  },
  "recent_models": {
    "large": [
      { "model": "glm-4.7", "provider": "nrp" },
      { "model": "gpt-oss", "provider": "nrp" }
    ]
  }
}
JSON

Verify it works by running:

crush run -m nrp/glm-4.7 "Hello"

You are now ready to use powerful open-source models for free in your terminal!

Alternatives

If you have access to GitHub Copilot, you can also use these tools with it. See Terminal-based AI Assistant with GitHub for Education.