AI Tools

Cherry Studio: The Complete Guide (2026)

Published 23 May 2026 • Updated 23 May 2026 • 8 min read

Quick answer. Cherry Studio is a free, open-source desktop client (Mac, Windows, Linux) that gives you one chat interface for OpenAI, Anthropic, Google Gemini, DeepSeek, Mistral, Qwen and local Ollama or LM Studio models. Install it from cherry-ai.com or GitHub Releases, paste your API keys in Settings → Model Providers, and you can switch between Claude Opus, GPT-5.5 and a local Llama model in the same conversation, with built-in RAG, MCP servers and document chat.

Updated 2026-05-23.

If you use more than one large language model in a normal week — Claude for code, GPT for brainstorming, Gemini for long-context document work, a local Ollama model for private notes — you have probably opened four browser tabs and three desktop apps to keep them straight. Cherry Studio collapses all of that into one desktop app. It is open source, has crossed 40k GitHub stars, and as of mid-2026 supports roughly 300 models across 50+ providers behind a single chat window.

This guide walks through what Cherry Studio is, how to install it on every supported OS, how to connect the major cloud providers and a local Ollama backend, how to build a private knowledge base, and how it compares to the obvious alternatives (LM Studio, Open WebUI, Msty, ChatGPT desktop).

What is Cherry Studio?

Cherry Studio is a cross-platform desktop client — not a model and not an inference engine. It is built on Electron and ships as a native installer for Windows, macOS (Intel and Apple Silicon) and Linux. Under the hood it speaks every common LLM API shape (OpenAI-compatible, Anthropic, Google AI Studio, Vertex, Ollama’s local API, Azure OpenAI, AWS Bedrock, plus dozens of region-specific providers).

The core idea: bring your own API keys (or run models locally), and Cherry Studio gives you one window with chat, document RAG, image generation, voice input, MCP-server tool use, agents, and a unified history. The app is licensed AGPL-3.0 for personal and team use, with a paid Enterprise edition for centralized admin, SSO and audit.

Why people use it

One UI for everything. Flip between Claude Opus 4.7, GPT-5.5, Gemini 2.5 Pro and a local qwen3:8b in the same chat.
No subscription stack. You pay providers directly via API instead of buying ChatGPT Plus + Claude Pro + Gemini Advanced.
Private RAG. Drop a folder of PDFs into a Knowledge Base; with a local embedding model your documents never leave the machine.
MCP support. Pre-wired filesystem, GitHub, web-search and memory MCP servers, plus the ability to add your own.
Open source. Source on GitHub at CherryHQ/cherry-studio; release notes are transparent.

Cherry Studio vs LM Studio, Msty, Open WebUI, ChatGPT desktop

Picking the right local-AI desktop is mostly about what mix of cloud + local you need.

App	Cloud LLMs	Local LLMs	RAG	MCP	Notes
Cherry Studio	50+ providers	Via Ollama / LM Studio	Built-in	Yes	Free, open source, every major OS
LM Studio	Limited	First-class (own engine)	Basic	Partial	Best for pure local inference
Msty	Most major providers	Via Ollama	Yes	Limited	Polished UI, freemium
Open WebUI	OpenAI-compatible	Via Ollama	Yes	Yes	Self-hosted web app, not native desktop
ChatGPT desktop	OpenAI only	No	Limited	No	Best if you only use OpenAI

If you only run local models, LM Studio is still the simplest path. If you need a single chat window across cloud + local with serious RAG, Cherry Studio is the strongest free option in 2026.

System requirements

Windows: Windows 10 or later, x64. ~500 MB disk for the app, more for embeddings.
macOS: macOS 11 Big Sur or later. Universal binary; separate Intel and Apple Silicon installers.
Linux: A recent x86_64 distro. AppImage, .deb and .rpm builds are published per release.
RAM: 4 GB free is enough for the app itself. Local model RAM is separate (e.g. an 8B local model needs ~6–8 GB).

How to install Cherry Studio

How to install Cherry Studio on Windows

Open cherry-ai.com or the GitHub releases page at github.com/CherryHQ/cherry-studio/releases.
Download the latest Cherry-Studio-Setup-x.y.z-x64.exe.
Run the installer. Pick per-user install unless you need it system-wide.
Launch Cherry Studio → Settings → General to set language and theme.

For a longer, screenshot-driven walkthrough see our Windows install guide.

How to install Cherry Studio on Mac

The cleanest path on Mac is Homebrew:

brew install --cask cherry-studio

Homebrew detects Intel vs Apple Silicon for you and makes updates a one-liner. If you prefer a direct download, grab the .dmg from GitHub Releases and drag Cherry Studio into /Applications. Detailed Mac walkthrough: Cherry Studio on Mac.

How to install Cherry Studio on Linux

Pick the artifact that matches your distro from the Releases page:

Ubuntu / Debian: download the .deb and run sudo apt install ./Cherry-Studio_x.y.z_amd64.deb.
Fedora / RHEL: sudo dnf install ./Cherry-Studio-x.y.z.x86_64.rpm.
Other distros: use the .AppImage: chmod +x Cherry-Studio-*.AppImage && ./Cherry-Studio-*.AppImage.

The AppImage path is the safest for distros without Cherry packaged in the repos.

Connecting LLM providers

Open Settings → Model Providers. Each provider has its own card with an API-key field and a list of models. Toggle a provider on, paste the key, and the provider’s models appear in the model picker at the top of every chat.

How to use Cherry Studio with Ollama

Install Ollama from ollama.com for your OS.
Pull a model: ollama pull qwen3:8b (or llama3.2, gemma4:e4b, deepseek-r1).
In Cherry Studio: Settings → Model Providers → Ollama.
Endpoint defaults to http://localhost:11434. Click Refresh Models; the models you pulled appear automatically.
Open a new chat, pick the Ollama model in the top picker, and you are running fully offline.

For a Windows-specific Cherry Studio + Ollama walkthrough see our Cherry Studio + Ollama on Windows guide.

How to use Cherry Studio with OpenAI

Create an API key at platform.openai.com/api-keys.
Settings → Model Providers → OpenAI, paste the key.
Enable the models you actually want (GPT-5.5, GPT-5-mini, o3, o3-mini). Disable the rest to keep the picker clean.
Optional: set a custom base URL if you proxy OpenAI calls (Helicone, Portkey, OpenRouter all work).

How to use Cherry Studio with Anthropic (Claude)

Create a key at console.anthropic.com.
Settings → Model Providers → Anthropic, paste the key.
Enable Claude Opus 4.7, Claude Sonnet 4.6, Claude Haiku as needed.
For a deeper look at Claude Opus capabilities, see our Claude Opus 4.7 guide.

How to use Cherry Studio with DeepSeek

Get a key at platform.deepseek.com.
Settings → Model Providers → DeepSeek, paste the key.
Enable deepseek-v4-chat and deepseek-r1 (reasoning). For a deep dive see our DeepSeek V4 guide.

How to use Cherry Studio with Google Gemini

Get a key at aistudio.google.com/apikey.
Settings → Model Providers → Google, paste the key.
Gemini 2.5 Pro is the strongest current model; Gemini 2.5 Flash is the cheap, fast tier.

How to build a knowledge base in Cherry Studio (local RAG)

Cherry Studio’s Knowledge Base feature ingests documents, chunks and embeds them, and retrieves the relevant pieces on every chat turn. With a local embedder, none of your documents leave the machine.

Choose an embedding model. The simplest option is nomic-embed-text via Ollama: ollama pull nomic-embed-text.
In Cherry Studio: Settings → Model Providers → Ollama, refresh; nomic-embed-text appears.
Open the Knowledge tab in the sidebar, create a new knowledge base, pick nomic-embed-text as the embedder.
Drag in a folder of PDFs, Markdown files, Word docs, or a list of URLs.
Open a chat, attach the knowledge base, ask questions. The retrieved chunks appear under each answer.

For larger corpora, swap in a cloud embedder (e.g. OpenAI text-embedding-3-large) for higher recall — at the cost of your documents being sent to that provider during indexing.

How to configure MCP servers in Cherry Studio

MCP (Model Context Protocol) lets the model call external tools — read files, hit APIs, query a database. Cherry Studio ships several connectors pre-wired (filesystem, web search, GitHub, fetch, memory) and supports custom ones.

Open Settings → MCP Servers.
Click Install on a built-in server (e.g. filesystem). Cherry downloads and registers it.
For a custom server, click Add and pick the transport (stdio command or HTTP URL).
In a chat, toggle the MCP server on under the message box; the model can now call its tools.

If the built-in install fails (GitHub downloads sometimes time out), grab the server manually from its repo and point Cherry at the local binary.

Agents and assistants

Cherry Studio includes a library of 300+ pre-built assistants (system-prompt presets) and lets you create your own. An assistant is just a chat preset: name, system prompt, default model, default temperature, attached knowledge base, attached MCP servers. They are the right level of abstraction for “a chat that always uses Claude Opus, with the company handbook attached, for product questions.”

For more on the agent stack across tools, see our AI coding agents guide.

Privacy and security

All app data — API keys, chat history, knowledge bases — is stored locally.
Chats hit the providers you enable. No Cherry-Studio-controlled relay.
With Ollama + a local embedder, you get a fully offline path.
The Enterprise edition adds SSO, audit logs and centralized provider management; the free edition does not.

Common issues and fixes

Ollama models do not appear. Confirm Ollama is running (curl http://localhost:11434/api/tags) and click Refresh Models in Cherry.
Knowledge base produces empty answers. The chunks did not retrieve anything — check the embedder is set and the documents finished indexing.
MCP install hangs. The download path goes through GitHub. Use a proxy or install the server binary manually.
Apple Silicon Macs flag the app as unverified. Right-click the app → Open the first time, or install via Homebrew which handles the quarantine.

FAQ

Is Cherry Studio free?

Yes. The desktop app is AGPL-3.0 licensed and free for personal and team use. You only pay the providers whose API keys you plug in (or run models locally and pay nothing). A paid Enterprise edition exists for organizations that need SSO, audit logs and centralized administration.

Does Cherry Studio work on Mac, Windows and Linux?

Yes. Native installers exist for Windows 10+ (x64), macOS 11+ (Intel and Apple Silicon as separate builds, or a universal Homebrew cask) and Linux (.deb, .rpm and .AppImage).

Can Cherry Studio run models locally without internet?

Yes. Install Ollama (or LM Studio), pull a model like qwen3:8b or gemma4:e4b, point Cherry at http://localhost:11434 in Model Providers, and you can chat fully offline. Pair it with a local embedding model like nomic-embed-text for offline RAG.

How does Cherry Studio compare to LM Studio?

LM Studio is best for pure local inference — it ships its own llama.cpp-based engine and a polished model browser. Cherry Studio is best when you want one chat window across many cloud providers and a local backend, plus built-in RAG and MCP. Many people run both: LM Studio (or Ollama) as the local engine, Cherry Studio as the UI.

Does Cherry Studio support MCP servers?

Yes. The MCP Servers settings tab supports both built-in installers (filesystem, GitHub, web search, fetch, memory) and custom stdio or HTTP servers. Toggle them on per chat under the message box.

What is the difference between Cherry Studio and Cherry AI?

They are the same product. cherry-ai.com is the official marketing site; the project name is Cherry Studio, the GitHub org is CherryHQ.

Does Cherry Studio store my API keys in the cloud?

No. API keys, chat history, attachments and knowledge bases are stored locally on your machine. Requests go directly from the app to each provider’s API. There is no Cherry-operated relay or proxy.

Can Cherry Studio do RAG over my PDFs?

Yes. Create a Knowledge Base, choose an embedding model (local via Ollama’s nomic-embed-text or cloud via OpenAI’s text-embedding-3-large), drag in your folder of PDFs/Markdown/Word/URLs, and attach the knowledge base to any chat. Retrieved chunks appear under each answer.

Which Cherry Studio version should I install?

Always grab the latest stable release from the GitHub Releases page or Homebrew (brew install --cask cherry-studio auto-updates). Pre-release builds are tagged on Releases for early features but expect rougher edges.

Is Cherry Studio open source?

Yes, AGPL-3.0. Source is at github.com/CherryHQ/cherry-studio. Anyone can audit, fork, or contribute.

Closing

Cherry Studio is the closest thing to a universal AI desktop in 2026: one window, every provider, local models, RAG, MCP, agents, and an open-source license. If you currently bounce between four AI tabs, install it for an afternoon and you will probably never go back.