Key takeaways
- ByteDance's Trae IDE introduces '@Agent' feature for creating custom AI agents.
- Custom agents can be tailored for specific development tasks, enhancing workflow efficiency.
- Users can define rules to customize agent behavior and coding style preferences.
- Trae is currently free, supporting macOS and Windows, with Linux support upcoming.
- Model Context Protocol (MCP) integration enhances connectivity between AI tools.
Trae, the AI-powered IDE developed by ByteDance, recently announced a range of updates to the tool, including the ‘@Agent’ feature that allows users to create custom agents.
Among the updates, the ability to create a custom agent takes the spotlight. TikTok's parent company mentions that the feature transforms the IDE into a versatile team of AI specialists. “Each Agent possesses unique capabilities designed for specific development tasks, allowing you to call upon the right Agent for their specific task,” the company highlighted in the blog post.
One can choose to create custom agents or use Trae’s built-in agents, such as Builder. To customise each agent, users can define rules that would affect the agent's behaviour.
The IDE is available for free at the moment with support for macOS, and Windows, with Linux support on the way.
📏Rules that can shape your agent's behavior. Build custom agents by combining Rules + MCP tools for more flexibility and collaboration. pic.twitter.com/khFI813xHk
— Trae (@Trae_ai) April 21, 2025
For instance, one can adjust the coding style preferences to accommodate complexity and adhere to architectural guidelines. The customisation feature ensures that users can create agents tailored to their workflows and promote human-AI collaboration for optimal results.
Additionally, Trae has added Model Context Protocol (MCP) support, which enables connections between Trae’s native agents and third-party extensions.
Along with these feature updates, the company has combined its chat to be used with its native agent, Builder. “In the previous version, switching between Chat (conversational programming) and Builder (AI-driven project development) often results in context loss and redundant explanations,” the blog read, “The new experience preserves your complete context as you move between conversational programming and project building”.
The company also clarified their reasoning behind the symbol as a prefix to the feature name, ‘@Agent’. It encourages users to interact with the agent in the same way they would mention a teammate on social media platforms.
