# OpenCode Integration with OfoxAI Configuration Guide

Source: https://ofox.ai/docs/integrations/opencode

[OpenCode](https://opencode.ai?utm=ofox_ai) is an open-source AI coding agent with a terminal TUI, a desktop app and IDE extensions. **Ofox is built into OpenCode's provider catalog** (sourced from [models.dev](https://models.dev/providers/ofox/?utm=ofox_ai)), so there is no custom Base URL to set up — select Ofox, paste your API key, and you can code with GPT, Claude, Gemini, DeepSeek and more.

## Prerequisites

- An OfoxAI account and API key ([get one here](https://app.ofox.ai/manage/api-keys))
- OpenCode installed:

```bash
# Pick one
curl -fsSL https://opencode.ai/install | bash
npm i -g opencode-ai@latest
brew install anomalyco/tap/opencode
```

The desktop app is available at [opencode.ai/download](https://opencode.ai/download?utm=ofox_ai).

## Option 1: Desktop App

### 1. Open provider settings

Open OpenCode, click **Settings → Providers** in the left sidebar, then click **Show more providers** at the bottom of the "Popular providers" list.

![Settings → Providers](https://ofox.ai/screenshots/en/opencode/01-providers.png)

### 2. Search for and select Ofox

Type `ofox` in the search box and click **Ofox**.

![Search for Ofox](https://ofox.ai/screenshots/en/opencode/02-search-ofox.png)

### 3. Enter your API key

Paste your OfoxAI API key and click **Continue**. Once connected, Ofox appears under "Connected providers".

![Enter your Ofox API key](https://ofox.ai/screenshots/en/opencode/03-api-key.png)

![Ofox listed under Connected providers](https://ofox.ai/screenshots/en/opencode/03-connected.png)

### 4. Enable Ofox models

Go to **Settings → Models**, search for `ofox` and switch on the models you want to use. Only enabled models show up in the session model picker.

![Settings → Models, enabling Ofox models](https://ofox.ai/screenshots/en/opencode/04-models.png)

### 5. Pick a model and start chatting

Create a new session, open the model picker below the input box and choose a model under the **Ofox** group. Click **Manage models** at the bottom to go back to step 4 at any time.

![Choosing a model under the Ofox group](https://ofox.ai/screenshots/en/opencode/05-select-model.png)

## Option 2: Terminal TUI

Start OpenCode in your project directory and type `/connect`:

```bash
opencode
```

```text
/connect
```

Search for `Ofox` in the provider list, press Enter and paste your API key. Then type `/models` and pick a model under Ofox.

Credentials are stored in `~/.local/share/opencode/auth.json` and shared between the desktop app and the terminal, so you only need to set this up once.

You can also log in and verify entirely from the command line:

```bash
opencode auth login          # choose Ofox, paste your API key
opencode models ofox         # list the models available under Ofox
opencode run -m ofox/anthropic/claude-sonnet-5.5 "Hello, introduce yourself"
```

## Optional: Environment Variable and Config File

OpenCode reads the `OFOX_API_KEY` environment variable automatically — handy for CI, or if you'd rather not store the key in a local file:

```bash filename="~/.zshrc"
export OFOX_API_KEY=<your OFOXAI_API_KEY>
```

To make an Ofox model the default, add it to `opencode.json` in your project root or in `~/.config/opencode/`:

```json filename="opencode.json"
{
  "$schema": "https://opencode.ai/config.json",
  "model": "ofox/anthropic/claude-sonnet-5.5"
}
```

Model IDs follow the format `ofox/<OfoxAI model ID>`; see the [Model Marketplace](https://ofox.ai/models) for the full list.

## Protocols: Built-in Provider vs Custom Provider

With the built-in Ofox provider, **each model uses its own native protocol by default** — there is nothing to choose:

| Models | Protocol | Endpoint |
| --- | --- | --- |
| Claude (`anthropic/*`) | Anthropic native | `https://api.ofox.ai/anthropic/v1` |
| Gemini (`google/*`) | Gemini native | `https://api.ofox.ai/gemini/v1beta` |
| GPT (`openai/*`) | OpenAI | `https://api.ofox.ai/v1` |
| Everything else (DeepSeek, Qwen, GLM, …) | OpenAI-compatible | `https://api.ofox.ai/v1` |

If you need to **pick the protocol yourself** — for example, calling Claude over the OpenAI protocol — the built-in provider can't do that. Add a custom provider instead and set the Base URL, API key and models manually.

### Add a custom provider in the desktop app

Go to **Settings → Providers → Show more providers**, choose **Custom OpenAI-compatible provider**, fill in the fields below and click **Submit**:

| Field | Value |
| --- | --- |
| Provider ID | Anything, e.g. `ofox-openai` (must not clash with the built-in `ofox`) |
| Display name | Anything, e.g. `Ofox (OpenAI protocol)` |
| Base URL | `https://api.ofox.ai/v1` |
| API key | Your OfoxAI API key |
| Models | ID = the OfoxAI model ID (e.g. `anthropic/claude-sonnet-5.5`), name is up to you; click "Add model" to add more |

![Custom provider: calling Claude over the OpenAI protocol](https://ofox.ai/screenshots/en/opencode/06-custom-provider.png)

A new **Ofox (OpenAI protocol)** group then appears in the model picker; every model in it is called over the OpenAI-compatible protocol.

### Or define it in opencode.json

```json filename="opencode.json"
{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "ofox-openai": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Ofox (OpenAI protocol)",
      "options": {
        "baseURL": "https://api.ofox.ai/v1",
        "apiKey": "{env:OFOX_API_KEY}"
      },
      "models": {
        "anthropic/claude-sonnet-5.5": { "name": "Claude Sonnet 5.5" }
      }
    }
  }
}
```

Refer to the model as `ofox-openai/anthropic/claude-sonnet-5.5`.

## Recommended Models

See the [Model Marketplace](https://ofox.ai/models) for recommended models.