OpenCode Guide

OpenCode is an open-source AI coding assistant framework. It supports integration with multiple AI models, including code generation, modification, and review workflows.

Installation

Choose an installation method based on your workflow.

macOS / Linux

curl -fsSL https://opencode.ai/install | bash

Windows PowerShell

Using WSL is recommended. Install it with the macOS / Linux method above inside WSL.

Install OpenCode globally:

npm install -g opencode-ai

After installation, run opencode in a terminal to start it.

macOS / Linux

brew install anomalyco/tap/opencode

Besides WSL, you can also use these package managers:

Chocolatey

choco install opencode

Scoop

scoop install opencode

Using WSL is recommended for best compatibility.

Connect to Flux

After installation, choose one of the following methods to connect OpenCode to Flux.

Using CC-Switch is recommended for centralized configuration.

Steps:

1. Follow Create API Key to generate an API key.
2. Follow CC-Switch to configure a unified provider.
3. Restart OpenCode after configuration is complete.

Step 1: Create the config file

In your project directory, create an opencode.json file.

Step 2: Fill in the configuration

Copy the following content into opencode.json and replace YOUR_API_KEY with your Flux API key.

{
  "$schema": "https://opencode.ai/config.json",
  "model": "openai/gpt-5.6-sol",
  "small_model": "openai/gpt-5.6-luna",
  "provider": {
    "openai": {
      "options": {
        "baseURL": "https://api.tryflux.net/v1",
        "apiKey": "YOUR_API_KEY"
      }
    }
  }
}

OpenCode automatically discovers models through its built-in openai provider. Configure other platforms under their matching built-in provider in the same way. To access multiple groups, enable composite key when creating the API key.

Step 3: Start OpenCode

Run this from the project directory:

opencode

Then run:

/init

Verify the Setup

Either configuration path can be confirmed with:

opencode models
opencode run -m openai/<model-id> "Reply with OK only"

opencode models lists the loaded models; seeing your model ID there means the configuration was picked up. opencode run makes a real, billed call, and a reply means the setup works.

If the commands fail or the model list is empty, first rule the client out with Test the Key and Endpoint on Their Own, then review opencode.json.

Related Pages

• Create API Key - pick a group and generate a key
• API Endpoints - address and protocol format
• Troubleshooting - locate a problem by symptom