Skip to main content

​
Overview

Quest Mode is an AI-assisted programming feature designed for complex, long-running development tasks. By describing your requirements in natural language, you can delegate tasks such as feature development, bug fixes, refactoring, and testing to the AI—significantly improving development efficiency. Quest Mode now supports both Local and Remote execution. Choose the option that best fits your needs.

​
Key Features

AI-Assisted Specifications (Specs) The AI assistant analyzes your project and automatically generates a detailed technical design document (Spec). This ensures a shared understanding of the task and aligns the AI’s execution with your goals. Autonomous Execution Once the Spec is completed or your requirements is clarified, then start to execute, the AI agent executes the task asynchronously. You can monitor progress in real time through the Action Flow view, which shows the execution plan, ongoing actions, and any issues encountered. AI-Assisted Task Report Upon completion, the AI generates a Task Report summarizing the changes, test results, and validation outcomes. This helps you quickly assess the quality and completeness of the work before deciding to accept or refine the results. Remote Delegation After finalizing your Spec design, you can delegate tasks for execution either locally or in the cloud, based on your needs. We recommend cloud execution for a superior experience, featuring intelligent scheduling, secure sandboxing, and a non-disruptive, native IDE workflow. This provides an on-demand and secure execution layer.

​
How to Create a Task

  1. Click New Task.
  2. Choose the relevant context (such as project files or code snippets) and enter your task description in natural language.
  3. Select a workflow
    • Spec First: Define the task specification before execution. Ideal for complex changes.
    • Direct Execute: Run commands immediately. Best for simple, quick tasks.
  4. Select a execution environment
    • Local: Run tasks directly on your machine, using git worktree to automatically isolate code changes and prevent conflicts from concurrent operations.Ideal for quick tests and immediate feedback. All file modifications are applied to your current project.
      Tips: We recommend configuring a startup script in Settings > Quest Mode to initialize the directory whenever Qoder creates a new worktree. This is ideal for installing dependencies and setting up your environment, ensuring every task starts in a consistent and ready-to-run state.
    • Remote: Delegate tasks to a powerful cloud environment. Perfect for long-running or resource-intensive jobs, keeping your local machine free and secure in a sandbox. Changes are automatically synced back to your IDE.
      Warning: Remote Mode requires a configured Git repository. Please set this up under Settings > Quest mode before proceeding. Currently, only GitHub is officially supported.
  5. Click Send or press Enter to start. The AI will begin drafting a detailed Spec based on your input or executing directly.

​
Technical Design

During the design phase, the AI:
  • Reviews your project’s memory and context
  • Analyzes the codebase structure
  • Generates a comprehensive Spec document
After the Spec is generated, you can edit it or refine it together with AI. It will be automatically saved in the.qoder/quest directory in your project. Once you’re satisfied with the Spec, click Start Now in the top-right corner to begin execution.

​
Action Flow

After the task starts, you’ll be taken to the Action Flow view, which displays:
  • The AI’s execution plan
  • Real-time output and logs
  • Progress status and any blocking issues
For simple tasks, the AI may skip generating a detailed plan and proceed directly to implementation. You can add new requirements at any time—even while the task is running—by sending messages in the chat box. The AI will adjust its plan and continue working.

​
Task Report

When the task is complete, the system switches to the Task Report view. The AI provides a structured summary, including:
  • Overview of implemented changes
  • Test results and validation outcomes
  • Detailed list of changed files
You can review each change individually and: In Local Mode
  • Click Accept to apply all proposed changes to your local files and archive the task.
  • Click Discard to revert all changes, restoring your files to their state before the task was run.
  • Click Finish to close the task without applying any changes. This is the primary option if no code was modified.
In Remote Mode
  • Click Checkout locally to download the remote branch to your machine for further review, local testing, or additional edits.
  • Click Create PR to open a pull request on your Git platform, initiating the formal code review and merge process.
  • Click Finish to close the task. This action is non-destructive; your remote branch and any created pull requests will be preserved.
If additional work is needed, return to the Action Flow view, send a new message, and the AI will continue from where it left off.