For the complete documentation index, see llms.txt. This page is also available as Markdown.

FAQ

Models and Connections

The model service has been filled with an API Key, so why can’t I chat?

Open [Settings] → [Model Services], confirm the provider is enabled, the API address is correct, and use the connection test. Also make sure [Settings] → [Default Model] is set to a model that is actually available under that provider. In enterprise proxy or firewall environments, also check the network proxy.

Why isn’t the model I want in the list?

First sync or manually add the model in Model Services, then check whether the model type matches the interface. Different pages filter models by capability; for example, Agent, drawing, and embedding models will not use exactly the same list.

What should I do when Agent prompts that the API gateway must be enabled?

Click [Enable and Start] in the prompt, or open [Settings] → [API Gateway] and start it manually. If the status is still abnormal, check whether the port is in use and whether local security software is blocking it.

Agent and Tools

Why can’t Agent see local files?

Confirm that the current task has selected the correct working directory and that the directory is still accessible. Knowledge base binding cannot replace the working directory; the knowledge base is used to retrieve processed materials, while the working directory is used to directly read and write files.

What should I do when Agent keeps waiting and doesn’t continue running?

Open [Status] on the right to see whether there are pending permission requests, background commands, subtasks, or provider retries. If you need more detailed information, enable Developer Mode and reproduce it once, then check the call chain.

I’ve already installed a skill, why doesn’t Agent use it?

Open the Agent edit window → [Skills] to confirm it is enabled, then send a new message. Installation and binding are two separate steps.

How do I get Agent to generate images?

First, in [Settings] → [Default Model] → [Drawing Model], select an available model, then go to the Agent edit window → [Built-in Tools] and confirm that [Generate Image] is enabled. Return to [Work] and simply describe the image.

MCP and Automation

The MCP server has started, but Agent still can’t find the tools. What should I do?

Confirm that the tools are visible in the server details, then check whether Agent [MCP] is bound. Configuration changes take effect from the next message; tools being turned off or permissions not approved will also prevent calls.

The channel shows connected but doesn’t reply. What should I do?

Check the bound Agent, workspace, allowed session ID, and channel permission mode, then review the channel logs. For platforms like Feishu and WeChat, also confirm robot permissions or the QR-code login status.

The scheduled task ran, so why didn’t I receive a message?

First check [Run History] to determine whether the Agent succeeded. If it succeeded but wasn’t delivered, check whether the channel has a usable receiving target; usually you need to send a message to the bot on the platform first.

Data and Windows

Will deleting a working directory in Cherry Studio delete the files on disk?

Deleting a working directory record from the Agent task list will not delete the directory on disk, but it will remove the task records under that directory. File operations during Agent execution may still modify real files, so check the permission mode.

How do I put a conversation or Agent task into a separate window?

Right-click the topic, task, or top tab, and choose [Open in New Window]. The independent window can be [Always on Top], and you can also click [Return to Main Window] to reattach it.

If you can’t find the corresponding issue, open [Settings] → [About Us] → [Feedback], and preferably choose [Submit with Agent]. Cherry Assistant will help check the current status, organize reproducible information, and ask you to confirm before submission.

Last updated

Was this helpful?