> For the complete documentation index, see [llms.txt](https://docs.cherryai.com.cn/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.cherryai.com.cn/docs/en-us/advanced-basic/agent-workspace/tools-knowledge-skills-mcp.md).
# Built-in Tools, Knowledge Base, Skills, and MCP
These four types of capabilities can all enhance the Agent, but they solve different problems: built-in tools handle common operations, knowledge bases limit retrievable materials, skills provide working methods, and MCP connects external tools and data.
### Differences and combinations
| Capability | Problems they solve | Example |
| -------------- | ----------------------------------------------------------- | ----------------------------------------------------------------------------------------- |
| Built-in tools | What the Agent can do directly in Cherry Studio | Read and write files, web search, generate images, memory, notifications, scheduled tasks |
| Knowledge base | What private materials can the Agent retrieve | Product specifications, contract templates, team handbook |
| Skill | What process and standards should the Agent follow | Weekly report format, code review checklist, writing guidelines |
| MCP | Which external tools and resources can the Agent connect to | Databases, browsers, third-party business systems |
Configuration path: left navigation \[Work] → open the Agent menu → \[Edit] → \[Built-in Tools].
### Built-in tools
Built-in tools are grouped by context, files, multimedia, orchestration, search, and terminal. Common capabilities include:
* \[Generate Image]: uses the configured drawing model;
* \[Knowledge Base Search] and \[Knowledge Base Management]: appear only after a knowledge base is bound;
* \[Web Search] and \[Web Scraping]: use the online capabilities configured in Cherry Studio;
* \[Memory]: save and retrieve stable information across tasks;
* \[Scheduled Tasks] and \[Notifications]: let the Agent manage plans and send results to channels;
* \[Task] and \[Workflow]: use sub-agents or orchestration to handle complex tasks;
* \[File to Markdown]: reads PDFs, Word, PowerPoint, Excel, OpenDocument, RTF, EPUB, and CSV files in the working directory or current session attachments, and passes structured content to the Agent for chunked processing;
* Session coordination tools: find existing Agent sessions, create new independent sessions, send tasks, and receive traceable completion results;
* File and Bash tools: read, create, modify files, and execute commands.
{% hint style="info" %}
\[File to Markdown] does not rewrite the original file, nor does it provide OCR for scanned PDFs. For scans, use \[OCR] or document processing capabilities first, then hand them to the Agent.
{% endhint %}
### Knowledge base scope
The knowledge base selected in \[Knowledge Base] is this Agent's retrieval boundary. If none is selected, the knowledge base tools will not load; if multiple are selected, the Agent will only work within those knowledge bases and will not automatically read other repositories.
### Skill
Enable already installed skills in \[Skills]. Skill content enters the Agent's working environment and is used to constrain the process and output. If you need a new skill, you can directly ask the Agent to help you find and install it in \[Work]; the manual management path is \[Settings] → \[Skills].
### MCP
First connect and start the server in \[Settings] → \[MCP], then return to the Agent editing window's \[MCP] binding. A server that is not running cannot be enabled; after connection, the tools it provides will enter the Agent's tool list and continue to be controlled by the permission mode. If the server also provides resources or prompt words, you can select and use them directly from the input area's \[+] panel.
{% hint style="danger" %}
Opening a tool does not mean you should skip approval. For file writes, terminal use, external data modifications, and potentially billable image generation, keep \[Confirm each time] by default; use looser permissions only for tools with trusted sources and clearly defined risks.
{% endhint %}
### Recommended combinations
| Task requires | Recommended combination | When to add more | Notes |
| ---------------------------------- | ---------------------------------------- | --------------------------------------------------- | ---------------------------------------------------------- |
| Process local files | File tools + dedicated working directory | Add skills when a fixed format is needed | Keep \[Confirm each time] first |
| Answer based on internal materials | Knowledge base + knowledge base search | After recall is stable, then hand it to the channel | First test unanswered questions |
| Repeatedly execute a fixed process | Skills + necessary built-in tools | Add MCP when external data is needed | Do not store account keys in skills |
| Call external systems | MCP + least-privilege credentials | Relax approval after the task stabilizes | Write, delete, and billable operations should be confirmed |
### User case: Marketing Research Agent
Bind the product materials knowledge base to the Agent, enable the "Research Report" skill, connect a trusted web retrieval MCP, and keep the built-in web scraping and file write tools. The Agent first determines the product scope from internal materials, then searches external information, and finally writes to the working directory according to the structure specified by the skill. Internal materials, external sources, and output format each have clear ownership, making later maintenance easier.
Why are there still no retrieval results after binding a knowledge base?
Confirm that the knowledge base has finished processing and contains retrievable content, then check whether the binding in the Agent editing window still exists. After a knowledge base is deleted, the related binding is automatically removed.
Do I need to reopen the task after enabling a skill?
No need to close the task. After editing the Agent, the new skill configuration will be loaded starting from the next message; the reply currently being generated will not be interrupted.
Do I need to reopen the task after enabling a skill?
No need to close the task. After editing the Agent, the new skill configuration will be loaded starting from the next message; the reply currently being generated will not be interrupted.
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.
Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:
```
GET https://docs.cherryai.com.cn/docs/en-us/advanced-basic/agent-workspace/tools-knowledge-skills-mcp.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.