fx × Vivgrid

fx

fx is an open-source coding agent for your terminal. Connect it to Vivgrid with a custom OpenAI-compatible model connection, then select among Vivgrid models in the same session.

Install fx

On macOS or Linux, install fx with:

curl -fsSL https://fx.sh/setup.sh | bash

Open a new terminal and check the installation:

fx --version

See the fx installation guide for other install options.

Configure fx to Use Vivgrid

Get an API key from the Vivgrid Console and set it in your shell:

export VIVGRID_API_KEY="viv-xxxxxxxxxxxxx"

Add a custom connection and model list to ~/.fx/settings.json. Merge these keys into your existing file rather than replacing other settings:

{
  "provider": "vivgrid",
  "providers": {
    "vivgrid": {
      "protocol": "openai-chat-completions",
      "base_url": "https://api.vivgrid.com/v1",
      "auth": {
        "type": "bearer",
        "env": "VIVGRID_API_KEY"
      },
      "model_metadata": {
        "gpt-6.1-sol": {
          "context_window": 1050000,
          "max_output_tokens": 128000,
          "supports_tool_use": true,
          "supports_vision": true
        },
        "claude-opus-5-5": {
          "context_window": 1000000,
          "max_output_tokens": 128000,
          "supports_tool_use": true,
          "supports_vision": true
        },
        "glm-5.3": {
          "context_window": 1000000,
          "max_output_tokens": 128000,
          "supports_tool_use": true
        }
      }
    }
  },
  "models": {
    "vivgrid": "gpt-6.1-sol"
  }
}

The model IDs and limits match the gpt-6.1-sol, claude-opus-5-5, and glm-5.3 catalog pages. fx uses the Chat Completions API for custom connections; Vivgrid supports these models on that API.

Verify the Connection

List the models declared in the connection:

fx models

Check the selected provider and model:

fx status

Start Using fx with Vivgrid

Start an interactive session with the configured default model:

fx

Choose another model with /model, or start a session with a specific model:

fx --provider vivgrid --model claude-opus-5-5
fx --provider vivgrid --model glm-5.3

You can also make a one-shot request:

fx ask --provider vivgrid --model gpt-6.1-sol "Explain the auth flow in this repository"

fx accepts an --effort option for models that support reasoning effort. Vivgrid's model-specific docs describe model/API compatibility, but do not specify how effort is handled on the Chat Completions surface. Treat effort as endpoint-dependent rather than assuming the custom connection forwards provider-specific reasoning controls.

Summary

A single Vivgrid key and https://api.vivgrid.com/v1 connection makes gpt-6.1-sol, claude-opus-5-5, and glm-5.3 available in fx. Configure the endpoint once, then select a model with /model or the --model option.