Overview

Qoder CN provides two AI Chat modes—Ask and Agent—to help solve coding problems, fix errors, debug, and troubleshoot runtime errors. Qoder CN supports multi-file edits, autonomous decision-making, codebase awareness, and tool use for end-to-end coding tasks. Qoder CN AI Chat core features:

• Multiple chat modes: Ask and Agent modes are available within a single conversation. Switch freely between modes based on the scenario and task complexity to maximize productivity.
• Automatic codebase awareness: Detects project frameworks, technology stacks, relevant code files, and error messages from task descriptions, without the need to manually add project context.
• Tool use: Uses 10+ built-in coding tools for file read/write, code search, and error diagnosis. Supports MCP tool configuration and integrates with ModelStudio MCP Square for flexible configuration.
• Terminal command execution: Autonomously decides which commands to run, generates them, and executes them in the terminal, significantly improving task execution efficiency.
• Project-level changes: Autonomously breaks down tasks and modifies multiple code files. Iterates through multi-turn chats or snapshot reverts to complete coding tasks collaboratively with Qoder CN.
• Memory awareness: Supports autonomous memory powered by large models. Throughout your conversations with Qoder CN, it gradually forms rich memories about you, your project, and your issues—the more you use it, the better it understands you.

​

Start a chat

​

Open the AI Chat panel

After installing Qoder CN in the IDE, click the Qoder CN icon in the sidebar or use keyboard shortcuts to open the AI Chat panel. Log on to start chatting. For more details, see Installation Guide.

​

Choose a chat mode

Qoder CN offers two chat modes within a single conversation. You can switch modes freely without starting a new chat. IDE support:

• Visual Studio Code: Ask and Agent modes are supported.
• Qoder CN IDE / JetBrains IDE plug-in: Ask and Agent modes are supported.

Select a mode based on your needs:

Ask	A pure coding Q&A mode. Choose Ask mode when you need code Q&A, knowledge Q&A, or coding help. Based on your question and provided context, it delivers solutions and suggestions without directly modifying project files.	See: Ask Mode
Agent	An autonomous coding task execution mode with autonomous decision-making, environment awareness, and tool use. Uses project search, file editing, and terminal tools to complete coding tasks end-to-end. Supports MCP tool configuration to fit your workflow.	See: Agent Mode

​

Input requirements

After choosing a mode, enter your requirements in the input box. Tips for effective descriptions:

• Structure your request: First, clarify what you need Qoder CN to help with. Include a clear goal, and describe the coding task and requirements step by step.
• Provide relevant context: Select code files, images, and other context so Qoder CN better understands your background and produces more accurate solutions.
• Specify expectations: Tell Qoder CN the language, conventions, format, and change goals you want it to follow. For example, "Generate English comments for each method when producing changes."
• Iterate step by step: Tell Qoder CN whether the suggestions meet your expectations and describe any specific flaws so it can improve. For complex tasks, break them into multiple requests and iterate with Qoder CN.

​

Code modification and review

​

Multi-file edits

In Agent mode, multiple project files may be modified. Each file goes through a generate-and-apply process. You can see the changed files and their status in the response card or workspace:

• Generating: Generates code snippet modification suggestions for each file based on the task breakdown.
• Applying: After suggestions are generated for a file, they are automatically combined with the original file content to produce a code change file (not the original file).
• Applied: The code change file is ready.

Click a file to see the generation process and diff for that file.

​

Review, accept, or reject modifications

Click the view changes button in the workspace or click a file in the file list to open the diff view. You can:

• Use the up/down arrows to navigate between change points in the current file.
• Click reject or accept on a change point to decide.
• Use the forward/backward arrows in the file-level controls to switch between changed files.
• Click reject or accept at the file level to decide.
• Edit the current change file locally.

After every code change file is accepted or rejected, the current snapshot enters one of these statuses:

• Accepted: A file with any accepted change is considered accepted. When all files are accepted, the current snapshot is Accepted.
• Partially accepted: A file with any accepted change is considered accepted. When only some files are accepted, the current snapshot is Partially accepted.
• Rejected: A file with no accepted changes is considered rejected. When all files are rejected, the current snapshot is Rejected.

​

Multi-turn iterations and snapshots

​

Refine requirements in multiple turns

In Agent mode, after a turn produces code change files, you can continue asking to add or adjust requirements. Qoder CN combines previous changes with the new requirements to produce one or more new code change files.

​

View snapshots and revert changes

To view or revert to changes from previous turns, click the drop-down arrow to see all snapshots produced in the current session. Select a snapshot to view its files and switch to it:

• The conversation automatically scrolls to the response card that produced the snapshot's code change files.
• The file list below the snapshot switches to the code change files of the selected snapshot. Click a file to view the change details.
• Click Switch to revert the current code changes to the state of the selected snapshot.

​

Start a new chat

While chatting in the panel, you can start a new chat in two ways: Method 1: Click the new chat button in the top right corner. The panel returns to its default state so you can start a new chat.

Method 2 (VS Code only): Type / in the input box and select /newChat. The panel returns to its default state so you can start a new chat.

​

View chat history

To review previous conversations, click the chat history icon () at the top right of the AI Chat panel. This shows the chat history for the current project. Click an entry to view the detailed conversation.

​

Context

Chats support rich context inputs, such as code files, directories, images, gitCommit, and knowledge bases, and allow you to freely combine context with your prompts to describe your requirements. For more information, see Context.

​

Memory

Qoder CN provides long-term memory. Throughout your conversations with Qoder CN, it gradually forms rich memories about you, your projects, and your issues, and automatically organizes and updates them. Memory helps Qoder CN interact with you more effectively—over time, Qoder CN understands you better and better. For more information, see Memory.

​

Tools

Qoder CN provides 10+ coding tools, including file search, file read, directory read, in-project semantic symbol search, file modification, error retrieval, and terminal execution. Qoder CN uses these tools autonomously, without requiring your confirmation or intervention. For more information, see Tools.

​

MCP

Qoder CN Agent mode supports MCP services. You can configure your own MCP services for the agent to use, extending the boundaries of the AI coding assistant and fitting your workflow more closely. It also integrates a third-party MCP service marketplace for one-click installation. For more information, see MCP.