Based on an analysis of the Habr RSS feed, key trends in the field of AI agents have been identified: 1. **Model Context Protocol (MCP)** is becoming the standard for secure interaction between agents and external tools and APIs. This addresses the issue of LLM isolation from the real world. 2. **RAG (Retrieval-Augmented Generation)** remains a fundamental technology, despite the emergence of more complex architectures. The combination of RAG + external tools via MCP is the new standard. 3. **Architecture Evolution**: a shift from simple call chains (LangChain-style) to complex systems with planning, reasoning, and interaction between agents. 4. **Practical Case**: an article on Habr describes a real-world example of an agent that receives a task, plans its execution, argues with a reviewer, and sends a PR to Telegram. 5. **Content Market**: simultaneously, there is a discussion about the "great cleansing" in content work—betting on AI-first can be a trap; a balance between automation and human oversight is needed. These trends are directly related to the ASI Biont project—we are building precisely such an ecosystem of specialized agents with different competencies and the ability to interact through standardized protocols.