We've improved our agent capabilities with better context understanding and greater autonomy. The direction we're heading with Agent 2.0 is accomplishing more complex tasks with maximum autonomy.
Here's a quick tl;dr on what's new:
-
Agents 2.0 now has longer context windows and acts more autonomously
-
One-click agent sharing with security protection
-
Zero-config setup for Go, Java, Python, Node.js
-
Available now for all users
Trae Agent 2.0: Enhanced Performance, More Reliable
Improved Context Window: Your Context Never Dies No more context amnesia mid-project. Trae Agent now maintains comprehensive session memory across extended conversations, remembering both historical details and critical project context with more precision. We automatically uses auto-compact when the context exceeds a certain capacity.
Your multi-turn conversations and complex coding tasks stay connected—Agents track key details across days-long development cycles like an experienced pair programming partner. One-click continuation with seamless context flow means less time re-explaining requirements and more time shipping code.
Agent Autonomy: Proactive Tool Selection Trae Agent intelligently anticipates your needs and automatically invokes the right tools without constant user input. We've enabled more proactive tool selection in chat to make the experience more "agentic". By using agentic context retrieval to autonomously gather and analyze context from your codebase, shell, and the Internet, we've minimized manual context provision, reducing your mental load.
The expanded autonomous tool selection significantly boosts task success rates. We've built AI that extends your thinking, not just executes your commands.
No More Proposal Bottlenecks We've removed the rigid proposal approval flow. Trae Agent now autonomously plans execution paths instead of requiring manual route planning.
Simplified decision-making reduces wait times and user intervention during development. We've optimized underlying infrastructure for faster response times, creating a more intelligent and user-friendly development workflow.
Share Agent: Collaborative Intelligence at Scale
Trae now supports one-click sharing and importing of custom Agents with built-in security checks.
You can publish your Agent to X (Twitter) or copy shareable links to paste anywhere—forums, docs, or direct messages. When sharing or importing, Trae automatically scans Agent and MCP configurations for sensitive information like API keys, tokens, and credentials to ensure safe knowledge exchange.
Import custom Agent into your Trae IDE with a shareable link—validated Agent and MCP configurations are added directly to your library.
This enables us to build toward an open Agent Ecosystem with security by default. Your expertise becomes community knowledge. Community knowledge becomes your toolkit.
Out-of-the-Box Language Support: No Config Required
Trae now provides native support for Go, Java, Python, and Node.js with zero manual setup required.
When you open a project, Trae automatically detects the language and guides you through installing the appropriate environment and extensions. No more wrestling with language-specific configurations—Trae handles the entire setup flow so you can focus on writing code.
Available Now
These feature updates are available now for all users—both free and Pro. Our Pro Plan is also available with more fast requests. View pricing