Concepts 101: Assistant / Agent / Skill / MCP / Channel

When you first open Cherry Studio and see terms like "Assistant", "Agent", "Skill", "MCP", and "Channel", you may feel confused — since they are all AI-related features, what exactly is the relationship between them?

This page explains it in plain languagethe definitions of each concept, the problems they solve, and the scenarios they are suited for, helping you build a complete understanding of the product.

Build the big picture at a glance

Think of Cherry Studio as a company:

• Model services (Provider) = the "energy company" that powers employees' brains (model service providers such as OpenAI, Anthropic, DeepSeek, etc.)
• Assistant = an AI colleague with a fixed persona AI colleague, such as a "rigorous code reviewer" or a "professional writing consultant"
• Agent (Cherry Agent) = an advanced colleague,with execution capabilities — able to autonomously read files, consult materials, run commands, and complete multi-step tasks
• Skill = a "professional capability pack" installed on a colleague, such as "can make PPTs" or "can draft patent applications", ready to use out of the box
• MCP = a "universal interface" for colleagues to connectexternal tools, which can connect to databases, Notion notes, GitHub, etc.
• Channel = dispatch a colleague toFeishu, WeChat, Telegram and other IM platforms to act as a group bot
• Scheduled tasks = set up aduty roster, and run automatically according to the schedule

Comparison table: when should you use which?

What you want to do	Recommended
Chat with AI and let itreply in a certain style(for example, "gentle and patient" or "professional and rigorous")	Assistant
Let AI read files on its own、execute commands on its own、complete multi-step tasks(for example, "help me organize this project's code and write documentation")	Agent (Cherry Agent)
Want to equip AI with aspecialized skill("can make Xiaohongshu image/text posts", "can draw flowcharts")	Skill
Let AI be able tocall external tools and services(query databases, scrape web pages, operate Notion / GitHub, etc.)	MCP
Let my AI work for colleagues in a Feishu group	Channel
Let AI Automatically at 9:00 every morningmake me a news briefing	Scheduled tasks(with Agent)

Detailed explanation item by item

1. Assistant — set a fixed "role" for AI

A normal conversation is like a "blank slate"; every time you open a new window, you need to re-explain to the AI, "you are a certain role, please reply in a certain style."

Assistant This set of "role settings"permanently saved, and can be used directly when reopened, avoiding repeated configuration. One assistant = one "system prompt + model parameter preset".

• How to create: in + the Launcher's assistant library , click + Create Assistant (fill in the name, prompt, and choose an Emoji), see assistant library.
• How to use: at the top of the assistant list on the left side of the main chat page, click+ Add Assistant, and start chatting after selecting one, see conversation.
• Feature layering: assistants can be attached to knowledge bases, skills, or MCP external tools.
• Detailed docs:assistant library

2. Agent — giving AI execution capabilities

In a normal conversation AI can only "offer suggestions" and cannot actually read files, run commands, or complete specific tasks.

Cherry Agent isable to autonomously execute tasksan advanced AI. You set a goal (for example, "organize all PDFs in this directory into Excel"), and it will automatically break down the steps, call tools, and execute step by step; when problems arise, it can also seek solutions on its own.

• How to useTop Tab Agent → Create an Agent
• PrerequisitesEnable API server + configure a Provider(such as CherryIN / Anthropic)
• Detailed docs:Agent

Core differences between Assistant and Cherry Agent:

• Assistant = only capable ofconversationa colleague
• Agent = canexecute tasksa colleague

3. Skill — installing an "AI professional capability app"

An agent itself is smart but is a "generalist". If you often handle a certain kind of specialized task (such as Xiaohongshu image/text posts, PPTs, financial analysis), you can installthe corresponding skillto enhance it.

Analogy:installing specialized apps on a phone. WeChat is for communication, Douyin for watching videos; skills are specialized software added to AI.

• How to use:Settings → Skills browse/install, and enable it in Assistant or Agent settings
• Source: built-in skills + community or official skill marketplace
• Detailed docs:Skill

4. MCP — AI's "universal interface"

Agents need to execute tasks, but AI itself doesn't have the ability to access external resources — How do you query the company database? How do you operate Notion / GitHub?

MCP (Model Context Protocol) is theunified interface standard. Developers build tool components according to this standard (called MCP Servers), and AI can then invoke these tools.

Analogy: the phone's USB / Type-C interface — chargers, USB drives, keyboards, and cameras can all connect because they follow the same interface standard. AI can likewise "connect to" various external tools through MCP.

• How to use:Settings → MCP Servers Add the needed MCP Servers (such as "query Notion", "operate GitHub", "access databases", etc.)
• Available objects: agents and (some) assistants can be attached to MCP
• Detailed docs:MCP tutorial

About reading local files: Cherry Studio already has built-in file upload, knowledge base, and other local file reading capabilities. Normal conversations and assistants can use them directly; MCP is not necessarily required. MCP is more suitable for extending external tools and services that Cherry Studio does not have built in.

Differences between Skills and MCP:

• Skill = built-in "AI professional capabilities" (how to write, draw, and format) — more on thesoftware
• MCP = letting AI connect to "external tools and data" (read files, query databases, call APIs) — more on thehardware

Analogy: skills are likeeducational background(decides the way of thinking), MCP is like atoolbox(decides the available resources).

5. Channel — connecting AI to IM groups

After an agent is configured, it can be inconvenient if you have to open Cherry Studio every time you want to chat.ChannelLet your agentstay in a Feishu group, and group members can trigger a response by @-mentioning it.

ChannelConnect Cherry Agent to IM platformsas a bot. Currently supported: Feishu / Telegram / QQ / WeChat / Discord / Slack.

• How to use: first create an Agent → Settings → Channel → choose a platform → fill in that platform's bot credentials
• Detailed docs:Channel

6. Scheduled tasks — AI's "duty schedule"

Suitable for periodic tasks such as "automatically generate yesterday's news briefing every morning at 9" or "summarize the team's workload every Monday".Scheduled tasksLet the Agent run automatically according to the schedule, without manual triggering.

• How to use:Settings → Scheduled Tasks → choose Agent + set trigger time + write task prompt
• Detailed docs:Scheduled tasks

Overall relationship (one sentence)

Provider Provide model capabilities → Assistant / Agent are colleagues with specific roles → throughSkillacquiring professional skills → through MCP connecting external tools → throughChannelbeing dispatched to IM platforms → throughScheduled tasksrunning automatically on schedule

Suggested learning path

For first-time use, it is recommended to proceed in the following order:

1. Step 1: in Settings → Model Services configure one Provider, completing the basic chat capability
2. Step 2: enter assistant library add assistants for common scenarios (such as "translation assistant") and start using them daily
3. Step 3Enable knowledge base, allowing AI to retrieve personal materials
4. Step 4: achieve automation → enable API server → configure Agent
5. Step 5: advanced capabilities → MCP connect tools / Channel connect IM / Scheduled tasks run periodically

No need to master everything at once; enable them gradually according to your actual needs.

---

💡 Get help and submit feedback

If you encounter any questions, bugs, or have suggestions for feature improvements during configuration or use, please refer to Feedback and Suggestions for the official channels provided.