Start a Conversation
Enter Your Request
Open a session, then enter your request in the input box at the bottom of the chat panel on the right
Select a Model
Click the model selector above the input box to switch between different models. Currently, all large language models provided by Interface AI that are compatible with the Anthropic API are supported.
Permission Modes
Cloud Code provides three permission modes that control how Claude Code performs operations:Accept Edits
Accept Edits
Claude Code requires your confirmation before modifying files. Suitable for scenarios where cautious operation is needed.Features:
- Manual approval is required for every file modification
- You can preview changes before deciding
- Suitable for important projects or for learning Claude Code’s behavior
Full Auto
Full Auto
Claude Code can automatically perform all operations. Suitable when you trust the AI to complete tasks independently.Features:
- No manual confirmation required; Claude Code executes directly
- Provides the highest development efficiency
- Suitable for rapid prototyping and for use after becoming familiar with Claude Code
Plan Mode
Plan Mode
Claude Code first creates a plan and waits for your approval before executing it. Suitable for step-by-step execution of complex tasks.Features:
- Claude Code first analyzes the task and creates an execution plan
- You can review and modify the plan
- After approval, it executes step by step according to the plan
- Suitable for developing large features
Upload Files
You can upload files for Claude to analyze:Click to Upload
Click the attachment icon next to the input box to select a file
Drag and Drop to Upload
Drag files directly into the input box area
- Code files (.py, .js, .ts, .html, .css, etc.)
- Text files (.txt, .md, .json, etc.)
- Image files (.png, .jpg, .gif, etc.)
Claude Code can analyze and understand uploaded content, including code logic and image content.
View Tool Calls
When Claude Code performs operations, you can view detailed tool call information:- Tool name (such as
edit_file,bash) - Input parameters
- Execution result
- Execution time
Tools Available to Claude Code
Claude Code can use the following tools in Cloud Code:| Tool | Description |
|---|---|
read_file | Read file content |
write_file | Create or overwrite a file |
edit_file | Edit an existing file |
bash | Execute terminal commands |
glob | Search for files |
grep | Search file content |
Control Responses
Stop a Response
There are two ways to stop Claude Code’s response:- Click the Stop button next to the input box
- Press the
Esckey
Claude Code will stop generating as soon as possible, but operations that have already been executed will not be rolled back.
Regenerate
If you are not satisfied with Claude Code’s response, you can:- Edit your message and send it again
- Provide more detailed instructions and ask Claude Code to try again
- Create a new session and start from scratch
Best Practices
Effective Prompts
Be Clear and Specific
Clearly describe the functionality and expected behavior you want
Provide Context
Tell Claude Code about the project background and tech stack
Request Step by Step
Break complex tasks into smaller steps and complete them one by one
Iterate with Feedback
Provide feedback based on the results so Claude Code can improve
Example: Effective vs. Ineffective Prompts
| Ineffective Prompt | Effective Prompt |
|---|---|
| ”Help me write code" | "Create a Python function that accepts a username and age, and returns a formatted welcome message" |
| "Fix the bug" | "A TypeError occurs on line 42 of login.py. Please analyze the cause and fix it" |
| "Optimize this" | "This function is slow when processing large arrays. Please optimize it with a more efficient algorithm” |
Related Documentation
Session Management
Manage session information in a project
File Management
Manage files in a project