What’s NEW in TRAE IDE

• Optimized some features.

• Updated the Max mode to support the Kimi-K2-0905 and GPT-5 models, and extend their context windows to a maximum of 200k and 272k tokens respectively.
• Supported enabling/disabling "Auto-Fix Lints" in the chat.

• Optimized some features.

• Supported setting custom sound effects for task status notifications.

• Optimized some features.

• Optimized the "Auto-Run" feature: After enabling "Auto-Run", all commands will run automatically, except for those added to the denylist by users, which must be executed manually.

• Supported adding problems as context via the # symbol.

• Optimized the interactive experience for cue.
• Fixed online issues.

• When creating Agents with MCP Servers, you can enable or disable tools in MCP Servers.

• Fixed some known performance issues.
• Updates to the SOLO mode:
  • Added OpenRouter for AI service.
  • Optimized some features.

• A new Max Mode has been added to the IDE mode to provide efficient and enhanced model capabilities for complex task scenarios such as long text analysis and multi-round complex sessions.
  • Max Mode provides model capabilities including up to 200 rounds of automatic tool calls and a maximum 200k context window.
  • Max Mode calculates fees based on Token consumption. The fees will be converted into fast request usage and then deducted from the remaining usages within your Pro plan or Extra Package. Max mode may lead to relatively high fast request consumption; please decide whether to enable it based on your actual needs.

• Fixed the issue of failed connection to some Streamable HTTP type MCP Servers.

• Added notifications for task completion and interruption for you to track the real-time status of tasks.
• Optimized the UI design for switching between Chat and Builder. Chat can now function as an agent that users can select via @.
• Updates to the SOLO mode:
  • Added payment service for you to quickly integrate the payment functionality into your projects.

• Supported automatically capturing the console log and fixing issues in SOLO Builder.
• Supported sending instructions during the AI's response, so as to calibrate the context.
• Improved the experience with Figma tool.
• Improved the experience with Supabase tool.

• Added the Privacy mode to enhance privacy protection. You can turn it on or off in account settings. For more information, refer to "Privacy mode".
• Added a task list to the IDE mode. You can view the tasks broken down by AI and track their completion status on the list.

• Added smart optimization for the input content to improve the accuracy of instructions.
• Updates to the SOLO mode:
  • Added a built-in Figma tool that automatically parses design files. You can convert design elements into runnable code with a single click.

• Supported quickly integrating the APIs of AI services, improving configuration efficiency.

• Supported deploying large-scale projects.

• Reduced power consumption and CPU usage in some cases.
• Optimized AI's reuse of resources from terminal during long chats.
• Optimized AI's execution of terminal commands in Windows environments.
• Fixed the issue where unaccepted code is overwritten when multiple editors are used simultaneously.
• Fixed the issue where the $ character interrupts chats.
• Updates to the SOLO mode:
  • Supported generating backend services by SOLO Builder.
  • Supported connection to Supabase projects, making user and data management more convenient.
  • Added the task list feature to display the tasks broken down by AI and the progress of each task.
  • Added editable technical architecture documentations that clearly presents the technical solutions given by AI.

• Introduced a brand-new visual design for a more immersive intelligent development experience.

• Launched the all-new SOLO Mode:
  • AI automatically analyzes and organizes your requirements, generating an editable product requirement document.
  • AI Intelligently orchestrates tools like the editor, browser, terminal, and doc viewer to adapt to various development scenarios.
  • Quickly build polished front-end applications with the SOLO Builder agent, enabling a streamlined end-to-end development workflow. For more information, refer to "SOLO Builder".
  • Deploy projects in one click and instantly generate a shareable link.
  • Tool panels automatically switch in real time to follow AI actions, clearly presenting each step and output of the development process.
  • AI automatically detects console errors within the application it builds.
  • The AI-built application can be edited by element.

• Supported voice input in AI chats. For more information, refer to "Voice".
• Supported generating random avatars for custom agents.
• Supposed adding the information from the Browser panel and Preview window as context for AI chats. For more information, refer to "Browser & preview".
• Added a dashboard for you to track your AI coding-related statistics. For more information, refer to "Dashboard".

• Added the intelligent programming tool - cue:
  • Optimized the capabilities of models, making the automatically completed code more accurate.
  • Supported multi-line edits, predicted edits, and jump to edits. For more information, refer to "Cue".

• Simplified the process of adding MCP servers to agents.
• Removed the built-in Vue-Official extension. You can install it from the extension store.
• Supported submitting subscription and payment issues in the Report Issue window.
• Fixed the issue of prompting "Output is too long, please enter \'continue\' to get more" in some scenarios.

• Introduced subscription plans and extra packages. Each plan provides different levels of access to AI models. You can now monitor your AI request usage in the Account Management Center. For more information, refer to "Plans & billing".

• Enhanced the AI chat experience. Improvements to the chat flow and quality include:
  • The fixed, pre-defined proposal in the Builder agent has been removed. The AI now responds dynamically based on context, enabling smarter and more adaptive interactions.
  • The AI can now autonomously invoke a wider range of tools, minimizing manual intervention and making development smoother and more efficient.
  • The IDE now better maintains context across multi-turn chats. You can continue Q&A sessions seamlessly using the “Continue” command, leading to a more natural and uninterrupted chat experience.

• Supported sharing your agents on social platforms. Users can click shared links to import agents into their Trae IDE, along with the associated prompts and MCP server configurations. For more information, refer to "Agent".
• Added environment setup guides for multiple programming languages, including Python, Java, Go, and more.

• Improved stability.
• Merged the Chat and Builder panels. Now you can access Builder as an agent using @Builder. For more information, refer to "Agent".
• Added #Doc and #Web as two types of context:
  • #Doc: You can add docs from URLs or by uploading .md / .txt files, and use them as context in chats. You can add up to 1,000 files with the total size being within 50MB.
  • #Web: You can paste URLs, and AI will automatically search the webpages and extract relevant content as context.

For more information, refer to "Context".

• Supported setting up rules to manage AI's behaviors, including:
  • User rules: Rules created on your personal preferences and will be applied to all projects.
  • Project rules: Rules created for one project and only work in that project.

For more information, refer to "Rules for AI".

• Improved the Agent capability:
  • Supported creating your own agents by customizing prompts and tools.
  • Provided two built-in agents: Builder and Builder with MCP.

• Supported enabling the Auto-Run capability for agents to make them automatically execute command and utilize tools. A denylist is also provided for you to add the commands that you want to block.

For more information, refer to "Agent".

• Supported Model Context Protocol (MCP):
  • Provided the MCP Marketplace for you to quickly access third-party MCP servers.
  • Supported adding MCP servers to agents to diversify their capabilities.

For more information, refer to "Model Context Protocol".

• Supported adding the AI models provided by OpenRouter as custom models. For more information, refer to "Models".
• Supported connecting to Windows Subsystem for Linux (WSL) to develop your project. For more information, refer to "Develop in WSL".
• Optimized the Preview extension by adding the Device Toolbar, which enables flexible page resizing and provides both responsive layouts and mobile-device-oriented adaptive layouts.
• Optimized the notification center by adding the "Do Not Disturb" mode and enabling message management by extension.

• Added Japanese as a user interface language.
• Supported installing themes from the Marketplace. For more information, refer to "Themes".
• Supported using AI to generate Git commit messages. For more information, refer to "Source control".
• Added notifications for excessive memory usage by extension during its runtime.
• Fixed the issue of port conflicts with IPv4 and IPv6 for the Webview extension.

• Supported adding custom models by using your own API keys. For more information, refer to Models.
• Supported remote development using SSH. Trae currently supports connections to remote hosts running Debian 11+ or Ubuntu 20+. Additional operating systems and versions will be supported in future updates. For more information, refer to Remote Development using SSH.

• Supported the 64-bit version of Windows.

• Enhanced the interaction experience and workflow of the Builder by better aligning it with default terminal behavior and refining the file deletion methods.
• Added the drag-and-drop functionality for the file tree, allowing users to drag files directly into the AI chat input box.
• Fixed commonly reported issues and made further improvements to the user experience.

• Provides the Builder mode that enables automated analysis, planning, execution of development requirements, as well as live preview.
• Supports generating project-level code through multi-round chat.
• Supports features such as fast apply, code review, diff view, and version rollback.
• Enables enhanced accuracy with rich context types like code snippets, files, folders, workspace, and terminal log.
• Supports adding images to session commands, such as screenshots, design drafts, to more efficiently express requirements.