What is Cherry Studio

Cherry Studio is an open-source AI desktop client that bills itself as an all-in-one AI workstation. It does not provide a model of its own. It is a single front end: add API keys for providers such as OpenAI, Anthropic, Gemini, DeepSeek, Kimi or Zhipu, and you can use all of their models from one window. Local models served by tools like Ollama work too.

On top of that, it packs multi-model chat, agents, knowledge bases, image generation, translation and notes into one app. It runs on Windows, macOS and Linux, stores data locally by default, and with a local model can work entirely offline.

The project is released under the GNU AGPL v3, so the source is public and you can customise and extend it. It is a practical choice for anyone who uses several providers and is tired of juggling browser tabs.

Compared with each vendor’s own web app, a client like Cherry Studio adds value in three ways. Aggregation: models from different vendors sit in one interface, so switching and comparing is easy. Control: chat history, knowledge bases and settings live on your machine rather than on one platform. Extension: MCP, skills, knowledge bases and scheduled tasks can be layered on top of any model instead of being limited by one vendor’s product. The trade-off is that you bring your own API keys, pay per use and spend a little time on setup. The project stresses that it is beginner-friendly, and the documentation is thorough enough that newcomers can usually get going by following it step by step.

Key features

Multi-model chat. Send one question to several models at once and compare the answers. Conversations are grouped by assistant and can be exported to Markdown, Word and other formats. Many preset assistants cover translation, coding, writing and more, and you can build your own. Markdown, maths formulas and live HTML previews all render inline.

Unified provider management. Built-in providers usually need nothing more than an API key, and “Fetch model list” pulls in what is available. A single provider can hold several keys used in rotation to ease rate limits. Third-party services that follow the OpenAI, Gemini or Anthropic API formats can be added as custom providers.

Agents and automation. On the Work page, agents can read files, run commands and complete multi-step tasks. You can give assistants and agents skills, connect external tools such as databases, Notion or GitHub through MCP, deploy an agent as a bot in Feishu, WeChat, Telegram or Discord, and schedule recurring runs such as a daily news briefing.

Local knowledge base. Import PDF, DOCX, PPTX, XLSX, TXT and MD files, or use URLs, sitemaps or typed-in text as sources. After import you can run test searches and inspect how content was chunked, and export a finished knowledge base to share.

Productivity tools. A quick assistant can be summoned from anywhere. The selection assistant shows a floating toolbar when you highlight text in any app, offering one-click translate, explain, summarise or polish. There is also an image panel, a translation panel, a built-in Markdown notebook, AI mini-apps and global search.

Backup and privacy. Local, WebDAV and scheduled backups are supported. Paired with local models, everything can stay on your machine.

How to use it

  1. Install. Download the package for your OS and chip from the official download page and follow the Windows, macOS or Linux guide.
  2. Add a provider. Open Settings → Model Services, choose a provider and enter your API key. Click “Fetch model list” and press + next to the models you want.
  3. Test and enable. Click “Check” to confirm the connection, then switch on the provider’s enable toggle in the top-right corner. Otherwise its models will not appear in the picker.
  4. Start chatting. Create an assistant or pick a preset, choose a model and ask. Select several models when you want to compare.
  5. Extend as needed. Build a knowledge base, configure MCP servers, turn on the selection assistant, and shape it into your own workspace.

One detail when entering an API address: if a provider gives you a full URL such as https://xxx.com/v1/chat/completions, enter only the root and Cherry Studio fills in the path. If the provider uses a non-standard path, enter the full address ending in # to stop it appending anything.

Use cases

  • Comparing models: ask several providers the same question and pick the best fit for your task.
  • Personal knowledge management: load papers, notes and product docs into a knowledge base and query them.
  • Translation and office work: translate selected text, summarise quickly and handle documents without switching apps.
  • Privacy-sensitive work: pair it with a local model to process sensitive material offline.
  • Light automation: handle routine jobs with agents, MCP and scheduled tasks.

FAQ

Is Cherry Studio free? The client itself is open source and free. Using third-party models means paying each provider’s API charges; local models have no API cost.

Does it come with a model? It is mainly a client, so you connect model services yourself. The docs also list free options such as CherryAI; check the documentation for which models are currently available.

I added a key but no models show up. Why? Usually one of two things: you did not fetch the model list and add models to it, or the provider’s enable toggle is still off.

What licence does it use? GNU AGPL v3. If you plan to modify it and offer it as a service to others, note the licence’s requirements around publishing source code.

Official website