# RikkaHub Integration with OfoxAI Configuration Guide

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

[RikkaHub](https://rikka-ai.com) is an open-source native Android AI chat client with a Material You design. It supports custom OpenAI / Google / Anthropic compatible endpoints, and ships with MCP, web search, memory, multimodal input, and more.

## Prerequisites

- A registered OfoxAI account with an API Key ([Get one here](https://app.ofox.ai))
- RikkaHub installed on your Android phone ([Download from the official site](https://rikka-ai.com/download) or [Google Play](https://play.google.com/store/apps/details?id=me.rerere.rikkahub))

## Choose a Protocol

RikkaHub has three provider types: OpenAI, Google, and Claude. All three work with OfoxAI:

| Provider Type | Suitable Models | API Base Url |
|-----------|---------|-------------|
| **OpenAI** (general) | All models, e.g. `openai/gpt-6-luna`, `deepseek/deepseek-v4.1-flash` | `https://api.ofox.ai/v1` |
| **Claude** (recommended for Claude models) | `anthropic/claude-sonnet-5.5`, `anthropic/claude-opus-5.5`, etc. | `https://api.ofox.ai/anthropic/v1` |
| **Google** (recommended for Gemini models) | `google/gemini-3.8-flash`, etc. | `https://api.ofox.ai/gemini/v1beta` |

If you are in mainland China, you can switch to the mirror domain `api.ofox.io`. The paths stay the same and the API Key works on both. The screenshots below use the mirror address.

## Configure the OpenAI Protocol

### Step 1: Open Settings

Open the RikkaHub left sidebar and tap the **Settings** icon in the bottom-right corner.

![Settings icon at the bottom of the sidebar](https://ofox.ai/screenshots/rikkahub/01-open-settings.jpeg)

### Step 2: Go to Providers

On the settings page, under the **Models & Services** group, tap **Providers**.

![Settings → Models & Services → Providers](https://ofox.ai/screenshots/rikkahub/02-settings-provider.jpeg)

### Step 3: Add a Provider

Tap the **+** in the top-right corner to create a new provider.

![The + button at the top right of the provider list](https://ofox.ai/screenshots/rikkahub/03-add-provider.jpeg)

### Step 4: Fill in the Configuration

Select **OpenAI** as the type, fill in the following, then tap **Save**:

| Field | Value |
|--------|-----|
| **Name** | `ofox` (or any name) |
| **API Key** | Your OfoxAI API Key |
| **API Base Url** | `https://api.ofox.ai/v1` |
| **Enable** | On |
| **Response API** | Recommended when using OpenAI models; the API path automatically becomes `/responses` |

![OpenAI-type provider configuration page](https://ofox.ai/screenshots/rikkahub/04-openai-config.jpeg)

RikkaHub concatenates **API Base Url** and **API Path** directly into the request URL. Do not add a trailing slash to the Base Url, and do not include `/chat/completions`. With Response API off, the API Path defaults to `/chat/completions` and can call every model on OfoxAI; with it on, requests go through the [Responses API](https://ofox.ai/docs/api/openai/responses) that OfoxAI recommends, which gives better native Prompt Caching.

### Step 5: Add Models

Switch to the **Models** tab at the bottom. RikkaHub automatically fetches the OfoxAI model catalog, so you can tick models directly from the **Avaliable Models** list. You can also tap **Add New Model** and enter a model ID manually, for example `openai/gpt-6-luna`.

Model IDs need the vendor prefix. Write `openai/gpt-6-luna`, not the bare `gpt-6-luna`. Browse the full catalog in the [Model Marketplace](https://ofox.ai/models).

### Step 6: Test the Connection

Go back to the **Configuration** tab, tap the plug icon at the bottom left to open **Test Connection**, pick a model, then tap **Test**. When **Non-streaming**, **Streaming**, and **Tool Call** all show a green check mark, the configuration works.

![Test Connection: non-streaming, streaming, and tool call all passed](https://ofox.ai/screenshots/rikkahub/05-openai-test.jpeg)

## Configure the Claude Protocol (Recommended for Claude Models)

For Claude models, we recommend the native Anthropic protocol, which gives you the full feature set and native caching benefits.

### Step 1: Create a Claude-Type Provider

Repeat Steps 1–3 above to create another provider (for example, named `ofox-claude`), so the OpenAI and Claude protocols can be used side by side. Select **Claude** as the type, fill in the following, then tap **Save**:

| Field | Value |
|--------|-----|
| **Name** | `ofox-claude` (or any name) |
| **API Key** | Your OfoxAI API Key |
| **API Base Url** | `https://api.ofox.ai/anthropic/v1` |
| **Enable** | On |
| **Prompt caching (cache_control)** | Optional. Turn it on to save input cost in long conversations. See [Prompt Caching](https://ofox.ai/docs/develop/advanced/prompt-caching) |

![Claude-type provider configuration page](https://ofox.ai/screenshots/rikkahub/06-claude-config.jpeg)

The Base Url for the Claude type **must include `/v1`**. RikkaHub appends `/messages` automatically, so leaving out `/v1` sends the request to `/anthropic/messages` and returns 404.

### Step 2: Add Models and Test

On the **Models** tab, add a Claude model (for example `anthropic/claude-sonnet-5.5`), then go back to the **Configuration** tab and tap the plug icon to test the connection. All three items should show a green check mark.

![Claude protocol connection test passed](https://ofox.ai/screenshots/rikkahub/07-claude-test.jpeg)

## Configure the Gemini Protocol (Recommended for Gemini Models)

For Gemini models, we recommend the native Gemini protocol.

### Step 1: Create a Google-Type Provider

Repeat Steps 1–3 above to create another provider (for example, named `ofox-gemini`). Select **Google** as the type, fill in the following, then tap **Save**:

| Field | Value |
|--------|-----|
| **Name** | `ofox-gemini` (or any name) |
| **API Key** | Your OfoxAI API Key |
| **API Base Url** | `https://api.ofox.ai/gemini/v1beta` |
| **Vertex AI** | Off |

![Google-type provider configuration page](https://ofox.ai/screenshots/rikkahub/08-gemini-config.jpeg)

The Google type points at Google's official address by default. **Change the API Base Url** to `https://api.ofox.ai/gemini/v1beta` (with `/v1beta`, no trailing slash). RikkaHub appends `/models/<model ID>:generateContent` automatically. Keep the **Vertex AI** switch off.

### Step 2: Add Models and Test

On the **Models** tab, add a Gemini model (for example `google/gemini-3.8-flash`, including the `google/` prefix), then go back to the **Configuration** tab and tap the plug icon to test the connection. All three items should show a green check mark.

![Gemini protocol connection test passed](https://ofox.ai/screenshots/rikkahub/09-gemini-test.jpeg)

## Recommended Models

| Model ID | Recommended Provider Type | Description |
|---------|--------------|------|
| `openai/gpt-6-luna` | OpenAI | The fastest and cheapest model in the GPT-6 family, good for everyday chat |
| `anthropic/claude-sonnet-5.5` | Claude | Claude's mainstay model with 1M context and strong all-round ability |
| `google/gemini-3.8-flash` | Google | Multimodal and fast, good for frequent use |
| `deepseek/deepseek-v4.1-flash` | OpenAI | Great value with strong Chinese performance |

## Getting Started

Go back to the chat page, find the OfoxAI provider you just added in the model selector, and pick a model to start chatting. RikkaHub's MCP, web search, multimodal input, and other features all work with OfoxAI models.

## FAQ

**Q: No model to choose from during Test Connection**

Test Connection can only pick from models already added to the current provider. Add at least one model on the **Models** tab first, then go back to the **Configuration** page to test.

**Q: The Claude-type provider test returns 404**

Check whether the API Base Url was written as `https://api.ofox.ai/anthropic`, missing the trailing `/v1`. The correct value is `https://api.ofox.ai/anthropic/v1`.

**Q: "Invalid API Key" error**

Make sure the API Key was copied in full from the [OfoxAI Console](https://app.ofox.ai), with no extra spaces before or after.

**Q: "Model does not exist" error**

The model ID is missing the vendor prefix. OfoxAI model IDs use the format `vendor/model-name`, for example `openai/gpt-6-luna`, `anthropic/claude-sonnet-5.5`.

**Q: Connection fails in mainland China**

Switch the API Base Url to the mirror address: `https://api.ofox.io/v1` for the OpenAI type, `https://api.ofox.io/anthropic/v1` for the Claude type. The API Key works on both.