How to Use Chatbox: Connect OpenAI, Claude, and Ollama in an Open-Source AI Desktop Client

Sat, 27 Jun 2026 09:32:07 +0800

Chatbox is an open-source AI desktop client. Its GitHub repository is chatboxai/chatbox.

Its positioning is simple: connect different AI models to one chat window. You can use it with OpenAI, Claude, Gemini, DeepSeek, Ollama, and other services, and it runs on Windows, macOS, and Linux.

If you only use the ChatGPT web app, Chatbox may not be necessary.
But if you often switch between multiple models, API keys, local models, and cloud models, it is more like a unified workspace.

What Can Chatbox Do?

Chatbox mainly solves three problems.

First, it puts multiple AI services into one client. You do not need to jump between ChatGPT, Claude, Gemini, local Ollama, and different API consoles.

Second, it provides a desktop client experience. Compared with a purely web-based tool, it is more suitable for keeping open for long periods, managing conversations, saving history, and maintaining fixed workflows.

Third, it lowers the barrier for API users. Many model providers only offer APIs, which are not friendly to ordinary users. Chatbox turns those APIs into an interface similar to a chat app.

Simply put, it is not a new model. It is an AI chat client.

Supported Platforms and Models

According to the project information, Chatbox supports mainstream desktop systems:

Windows
macOS
Linux

It also supports many model sources. Common uses include:

OpenAI API
Claude / Anthropic
Google Gemini
DeepSeek
Ollama local models
Other OpenAI-compatible APIs

The most practical parts are OpenAI-compatible APIs and Ollama. Many model services now provide an interface similar to OpenAI’s /v1/chat/completions. As long as the base URL, API key, and model name match, they can usually be connected to this kind of client.

How to Install It

The safest way is to download it from the official repository or official website entry:

Open the Chatbox GitHub repository.
Go to the Releases page.
Choose the installer for Windows, macOS, or Linux.
Install and open Chatbox.
Add your model provider and API key in settings.

For ordinary users, building from source is not recommended at the beginning. Downloading a release package is easier.

If you use local Ollama, first confirm that Ollama is running:

`1`	`ollama list`

Then add the local endpoint in Chatbox. A common address is:

`1`	`http://localhost:11434`

Field names may change between versions, but the idea is mostly the same: choose a provider, fill in the endpoint, fill in the model name, save, and test.

Who Is It For?

Chatbox is suitable for four types of users.

The first type is API users. You already have keys for OpenAI, Claude, Gemini, DeepSeek, or another model service, but you do not want to build your own frontend.

The second type is local model users. You run local models such as qwen, llama, or deepseek with Ollama and want a convenient desktop chat interface.

The third type is multi-model comparison users. You often want to send the same question to different models and compare which answer is better.

The fourth type is people who prefer a desktop experience. You do not want to open a browser tab every time or scatter conversations across different websites.

How Is It Different From the ChatGPT Web App?

The ChatGPT web app is an official OpenAI product. You log in and use it directly, with a complete feature set. It is suitable for most ordinary users.

Chatbox is more like a universal shell. It does not come with a “universal model” by itself. Instead, you connect your own model services.

The difference can be understood this way:

ChatGPT web app: ready to use, suitable for users who only use OpenAI.
Chatbox: requires configuration, but can connect multiple models and APIs.
ChatGPT web app: features are controlled by OpenAI.
Chatbox: better for managing your own APIs, models, and local services.

If you only ask a few questions every day, the web app is simpler.
If you want to manage multiple models in one place, Chatbox is more suitable.

How Is It Different From Ollama and Open WebUI?

Ollama is a local model runtime. It is responsible for running models. It is not a complete desktop chat product by itself.

Open WebUI is a web-based local AI interface, often used together with Ollama. It is suitable for managing local models, users, knowledge bases, and related features in a browser.

Chatbox is more of a desktop client.

The division can be simple:

Ollama: runs local models.
Open WebUI: provides a browser-based local AI workspace.
Chatbox: provides desktop chat and multi-model connection.

If you only want to run one local model on your own computer, Ollama + Chatbox is simple enough.
If you need multi-user access, browser access, knowledge bases, and service-style management, Open WebUI is more suitable.

How Is It Different From Cherry Studio?

Chatbox and Cherry Studio both belong to the multi-model AI client category, and their user groups overlap.

In general, Cherry Studio emphasizes model-provider management, knowledge bases, Agents, MCP, and a more complete workflow experience. Chatbox is lighter and better suited to people who want simple desktop chat, model switching, and API access.

You can choose based on need:

Want stable chat and multi-model switching: Chatbox is more direct.
Want more workflows, knowledge bases, and plugin capabilities: compare Cherry Studio.
Already using one client: no need to migrate frequently; focus on conversation export, model support, and stability.

Things to Watch Out For

First, keep your API keys safe. Do not send them to others or share screenshots of your settings page.

Second, a third-party client sends content through the model services you configure. Do not send sensitive documents, internal company code, or customer data to uncertain model endpoints.

Third, billing is handled by the model provider. Chatbox is only the client; the tokens are consumed by OpenAI, Claude, Gemini, DeepSeek, or whichever service you connect.

Fourth, local models are not completely free. Running local models with Ollama does not consume cloud API fees, but it uses CPU, memory, VRAM, and electricity. Larger models require stronger hardware.

Fifth, open-source projects also require attention to updates. It is best to download installers from GitHub Releases or the official website, not from unknown file-sharing sites or repackaged sources.

My Take

The value of Chatbox is simplicity.

There are more and more AI tools now. The issue is not that no models are available, but that every model has its own entry point, price, account, and limits. Chatbox brings those entries into one desktop client, which is convenient for people who switch models often.

It is not meant to replace every professional workflow, and it may not have more features than official web apps.
But if you want to manage cloud APIs, local Ollama, and OpenAI-compatible endpoints in one window, Chatbox is an open-source option worth trying.

Chatbox on KnightLi Blog