Designed for Agent trajectory data synthesis, supports large-scale parallel execution, automatically records complete tool calls and environment interactions
🖥️
Multiple Resource Backends
Supports VM desktop automation, RAG retrieval, Bash CLI, Browser web automation, Code execution and more
📦
Session Management
Flexible session lifecycle with worker isolation, suitable for multi-process data synthesis scenarios
🌐
Standardized Interface
Unified HTTP API with standardized response format for easy data collection and post-processing
🔌
Extensible Architecture
Easy to develop new backends and tools with hot-plug support, quickly adapt to different synthesis needs