Complete Guide to Cherry Studio AI: The Command Center for Your AI Models

Using different applications to access ChatGPT, Claude, Gemini, and local templates can significantly fragment your work. Each service has its own interface, history, subscription, and document management methods, so constantly switching between platforms becomes impractical.

Cherry Studio brings together numerous artificial intelligence providers and models within a desktop applicationFrom a single interface you can configure cloud services, use models run with Ollama or LM Studio, create wizards, consult documents and connect external tools.

This guide will show you how to install Cherry Studio on Windows, add providers, use local models, and properly configure its main functions. We'll also explain what data stays on your computer and what data is sent when using cloud services.

What is Cherry Studio and what is it used for?

Cherry Studio is an AI client available for Windows, macOS, and Linux. Its Community Edition is open source and distributed under the Cherry license. GNU AGPL 3.0.

The application does not inherently include all business models. It primarily acts as an interface through which you can connect different services via their APIs. Its capabilities include:

• Configure providers such as OpenAI, Anthropic, Gemini, DeepSeek, and other compatible services.
• Use local models using Ollama or LM Studio.
• Hold conversations with different models from a single application.
• Compare the responses of several models.
• Create wizards with specific instructions, templates, and tools.
• Import documents and web pages into knowledge bases.
• Connect MCP servers and other external tools.
• Generate images using compatible providers.
• Save and organize conversations, topics, and files.

Current versions also include agent-based features capable of breaking down tasks into steps and using tools. The exact availability of each feature depends on the model, vendor, and configured permissions.

What you should know about Cherry Studio's privacy

Cherry Studio saves items such as chat history, preferences, configured keys, and knowledge base data to your computer by default. This provides more control than a purely web-based application, but This does not mean that all conversations are processed locally..

When you select a model from OpenAI, Anthropic, Gemini, CherryIN, or another remote provider, Cherry Studio must send that service the necessary prompt, history, attachments, or retrieved context needed to generate the response. The handling of this information will then depend on the terms and policies of the chosen provider.

To have a conversation entirely within your team, you need to use:

• A local model run using Ollama, LM Studio, or another compatible server.
• A local embedding model for knowledge bases.
• A local reranker, if the configuration uses one.
• MCP tools and servers that do not send information to external services.
• Local backups instead of synchronization via remote services.

Cherry Studio supports local backups and synchronization via WebDAV. While WebDAV is convenient for working across multiple computers, it means data is no longer stored exclusively on the device. Before using it with sensitive information, check where the WebDAV server is hosted and how the files are protected.

Requirements to install Cherry Studio on Windows

Cherry Studio offers different versions for x64 and ARM64 computers. Most computers with Intel or AMD processors use Windows x64ARM64 files are intended for computers with compatible ARM processors.

If you don't know what architecture your computer uses, open:

1. Configuration.
2. System.
3. About.
4. See the section System type.

On the versions page you can usually find two formats for Windows:

• Setup.exe: Install Cherry Studio on the system and create the corresponding shortcuts.
• Portable.exe: It allows you to run the application without performing a conventional installation.

To use cloud providers, you'll also need an internet connection and a valid API key. If you're running models locally, the RAM and graphics memory requirements will depend on the size and quantization of the chosen model.

How to download Cherry Studio for Windows

The easiest way to get the application is to use the Cherry Studio official websiteYou can also download it from the section of official GitHub versions.

Avoid downloading installers from third-party websites. Besides reducing the risk of using a compromised file, the official website allows you to locate the latest version and correctly select the architecture.

For most computers, you must download the file whose name includes:

x64-setup.exe

Choose the ARM64 version only if Windows indicates that your computer uses this architecture. If you want to try the application without installing it, you can download the variant that contains the word portable.

How to install Cherry Studio on Windows step by step

1. Open the installer downloaded from the official website or GitHub.
2. If Windows displays the User Account Control warning, check the file source before allowing it to run.
3. Follow the steps indicated by the installation wizard.
4. Finish the process and open Cherry Studio from the Start menu.
5. Select the language and review the application's initial preferences.
6. Access the settings to configure at least one provider and one model.

The interface may change between versions, especially after major updates. However, the model settings are located within the settings menu. Modeling services or the equivalent section dedicated to suppliers.

How to add OpenAI or Anthropic or Gemini

To use a commercial provider, you need to obtain an API key from their official platform. A subscription to ChatGPT, Claude, or Gemini does not necessarily include credits to use their API, as web access and the API are usually billed separately.

The general procedure in Cherry Studio is as follows:

1. Open the application settings.
2. Enter Modeling services.
3. Select the provider you want to configure.
4. Paste the API key in the corresponding field.
5. Check the API address. If you're using the official service, you should usually keep the default address.
6. Activate the provider.
7. Press the option to check the connection or get the list of models.
8. Select and enable the models you want to display in the application.

A valid key does not guarantee that your account will have access to all models. Availability may depend on the provider, region, balance, organization, and account permissions.

You can also add services compatible with OpenAI or Anthropic APIs using a custom provider. In that case, you must enter the base address provided by the service and select the correct API type.

How to connect Cherry Studio with Ollama

Ollama allows you to run models on your own computer. To use it with Cherry Studio, you must first install Ollama and download at least one compatible model.

For example, you can download a relatively lightweight model from a terminal using:

ollama pull qwen3:4b

Check the installed models with:

ollama list

Then open Cherry Studio and follow these steps:

1. Go into the settings of Modeling services.
2. Select Ollama.
3. Verify that the service address is http://localhost:11434.
4. Leave the API key empty if Ollama is running locally with its default configuration.
5. Get the list of available models.
6. Activate the model you want to use.
7. Select that model when starting a conversation.

The model continues to run using Ollama. Cherry Studio only provides the interface, organizes the conversations, and sends the requests to the local service.

How to use Cherry Studio with LM Studio

LM Studio can also expose local models using an OpenAI-compatible API. Before connecting, open LM Studio, load a model, and start your local server.

Next, create or select an OpenAI-compatible provider in Cherry Studio and enter the address displayed by LM Studio. An address similar to the following is typically used:

http://localhost:1234/v1

The port may change if you modified it in LM Studio. Some interfaces request an API key even if the local server doesn't validate it; in that case, you can use a test value, unless you have enabled authentication on the server.

How to compare several models in Cherry Studio

One of Cherry Studio's most useful features is simultaneous communication with multiple models. It allows you to send the same query to different systems and compare their responses without manually copying the prompt between multiple pages.

This function can be useful for:

• Compare the quality of different models when programming.
• Evaluate different versions or translations.
• Compare local and remote models.
• Check which one best follows certain instructions.
• Balancing quality, speed, and cost.

Please note that each response may generate a separate call. If you select three payment options, the prompt will be sent three times, and usage will be billed separately by each provider.

How to create a personalized assistant

Cherry Studio includes hundreds of pre-configured wizards and allows you to create your own. A wizard brings together a set of instructions, a predefined template, and, when necessary, knowledge bases or tools.

To create a new one:

1. Open the attendees section.
2. Create a new assistant.
3. Enter a name and a clear description.
4. Select the model you will use.
5. Write the system instructions.
6. It limits its function, tone, and response format.
7. Add a knowledge base or tools only if you need them.
8. Save the wizard and test it with several real-world examples.

The instructions should explain what the assistant should do, what information they can use, and what actions they should not take. A title like “Marketing Expert” provides less guidance than a detailed description of the audience, format, objectives, and constraints.

How to use files and knowledge bases

Knowledge bases allow you to import files and web pages to retrieve snippets related to a question. Cherry Studio supports various formats, including PDF, Word, Excel, text, and Markdown documents.

To create a knowledge base:

1. Open the section corresponding to the knowledge bases.
2. Create a new base.
3. Select a compatible embedding template.
4. Import the documents or pages you want to consult.
5. Wait for the application to process and index the content.
6. Associate the base with an assistant or conversation.
7. Ask specific questions and review the retrieved fragments.

The RAG function can better substantiate the answers, but It does not eliminate hallucinationsThe result depends on the quality of the document, the text extraction, the fragmentation, the embeddings, the retrieval, and the generative model.

Always verify the original sources when working with academic, medical, legal, financial, or business information. An answer generated in a few minutes is no substitute for a literature review or human verification.

How to use MCP tools safely

The Model Context Protocol allows Cherry Studio to connect to servers that provide external tools, resources, and sources. Depending on the server installed, the model could query a database, access specific files, browse the internet, or invoke an API.

MCP does not automatically turn any conversation into an autonomous agent. The chosen model must support tools, and Cherry Studio must have a wizard or agent configured that can use them.

Before installing an MCP server:

• Check who maintains the project and review their code when possible.
• Limit the folders and credentials that can be accessed.
• Do not grant write permissions if you only need to read information.
• Avoid providing general access to the file system.
• Review each sensitive action before authorizing it.
• Do not enter secrets directly into prompts or shared configurations.
• Deactivate the servers you are not using.

An MCP server can run processes or communicate with external services. Therefore, you should treat it like any other application with access to your computer, and not just as a visual extension.

Cherry Studio Community, Enterprise, CherryAI and CherryIN

The edition Cherry Studio Community It is free and open source under the AGPL-3.0 license. Commercial use is permitted provided the terms of that license are respected. Companies requiring different terms can consult a commercial license.

Cherry Studio Enterprise It is geared towards organizations that need centralized management of models, employee accounts, permissions, shared knowledge bases, backups, and deployment on their own servers or in a private cloud.

These two services should not be confused either:

• CherryAI: Integrated access to certain free models offered from the application, subject to its terms and conditions.
• CherryIN: provider or gateway from which different models can be configured using an account and an API key.

The models, promotions, limits, and prices of these services may change. Before using them frequently, check the updated cost and verify which company processes the requests.

Cherry Studio versus other artificial intelligence interfaces

Tool	Main focus	Local installation	Multiple suppliers	Knowledge bases
Cherry Studio	AI-powered desktop and workspace client	Yeah	Yeah	Yeah
ChatGPT	OpenAI Services and Applications	Client application; models hosted by OpenAI	Not as a primary function	Projects, files, and GPT according to the plan
LobeChat	Multi-model web and desktop interface	Yes, depending on the method chosen.	Yeah	Yeah
AnythingLLM	Workspaces and document consultation	Yeah	Yeah	Yeah

There is no single best option. Cherry Studio is particularly appealing to those seeking a desktop application with multiple vendors, local models, assistants, agents, and document management capabilities. AnythingLLM may be a better fit for certain workspace-centric deployments, while LobeChat offers various web, desktop, and self-hosted options.

Common mistakes when setting up Cherry Studio

The supplier does not show any models.

Verify that the API key is valid, the provider is activated, and that you have used the option to obtain or manage the model list. Some accounts do not have access to the entire catalog.

The API key works on the web, but not in Cherry Studio.

A subscription to a web application does not necessarily equate to API access. Verify that the key comes from the provider's developer platform and that the account has sufficient funds or that billing is enabled.

Cherry Studio does not detect Ollama

Verify that Ollama is running and that the configured address is:

http://localhost:11434

You can also check the service from a terminal:

ollama list

If Ollama is on another computer, you will need to use its network address and securely configure remote access.

The local model appears, but responds very slowly.

The model may be too large for the available RAM or graphics memory. Try a lighter quantization or a model with fewer parameters and check using [method name missing]. ollama ps how it is running.

The knowledge base cannot find the content

Verify that the documents have finished processing, that the embedding model is working, and that the database is associated with the correct wizard. Adjusting the content splitting and similarity threshold may also be necessary.

An MCP server fails to start

Some servers require Node.js, Python, environment variables, or additional commands. Check their logs and confirm that all dependencies are installed. On Windows, you should also verify that the executable is available in the PATH environment variable.

The responses generate more expenses than expected.

Check if you're sending the query to multiple models simultaneously, using an excessively long history, or attaching large documents. Each provider may charge separately for both incoming and outgoing tokens.

Is Cherry Studio worth installing?

Cherry Studio is an interesting alternative for centralizing cloud and on-premises models without relying on a single web interface. Its main advantage is bringing together conversations, providers, assistants, agents, documents, and tools within a single cross-platform application.

However, privacy depends on the configuration. Saving the history locally does not prevent prompts from being sent to the selected provider. If you want the processing to remain on your computer, use local models, embeddings, and tools.

To get started, simply install the application, configure a provider, and enable one or two models. Then you can gradually add assistants, knowledge bases, and MCP servers. This makes it easier to see which features you actually need and control both permissions and call costs.