Cloud.ru Evolution AI Agents is a service on the Cloud.ru cloud platform for creating autonomously operating AI agents. According to official materials, AI agents are capable of analyzing information, learning from data, planning actions, and adapting to changes in their operating environment.
Key capabilities
Multi-agent systems: Up to 5 different AI agents can be combined into a single unified system, coordinating their interactions through Agent-to-Agent technology to solve business tasks.
MCP protocol integration: Supports integration with MCP, flexibly connecting AI agents to different data sources; you can use self-built MCP or ready-made MCP from Маркетплейс (Marketplace).
Monitoring and tracing: Integrated with Arize Phoenix, it can trace data and request paths, and view how AI agents make decisions and how they process data and tools; request execution is monitored through OpenTelemetry tracing, and the system automatically analyzes answer accuracy and identifies unreliable data.
Enterprise knowledge base integration: Integrated with the Evolution Managed RAG service, it can connect to enterprise knowledge bases to obtain relevant and detailed answers, improving answer accuracy.
Large language model usage: You can choose ready-made public models billed by token in the Evolution Foundation Models service, or run your own models in the Evolution ML Inference service.
Development tool integration: Provides development flexibility based on the Agent Development Kit (ADK), allowing you to create, deploy, and manage agent architectures through development tools such as Langchain, ADK, OpenAI API, and SmolAgent.
AI Workflows: Provides a platform for creating automated workflows without programming.
Evoclaw autonomous agent: An open-source personal assistant that combines neural network agent capabilities and can integrate with external services as well as major operating systems.
How to use
Official materials state that AI agents can be created in three ways:
- Ready-made solutions: No programming skills required, add AI agents from Маркетплейс;
- Own infrastructure: Use existing Docker images;
- Create from scratch: Create AI agents based on official templates.
AI agents learn new skills or connect data through MCP servers and system prompts (system prompt) Once configuration is complete, use understandable instructions (prompts) to tell the AI agent how to access new information or services and how to process them. Integration with external systems is provided by MCP servers, and integrable systems (such as enterprise instant messaging tools and task trackers) are published on Docker Hub.
Agent Space is a desktop and mobile application for convenient interaction with AI agents and multi-agent systems, supporting Windows / Linux, allowing users to use agents to handle complex non-deterministic tasks without writing code or using technical interfaces.
Applicable scenarios
The officially listed AI agent examples include: virtual assistants, chatbots, intelligent recommendation systems, and business process automation systems. The service is positioned as running AI agents and multi-agent systems in the cloud, automating complex business processes, optimizing costs, and improving customer experience.
Other features
Officially, the service provides a unified platform for handling models, reducing the time and resource costs of integrating scattered components and accelerating the creation of AI solutions; the interface is intuitive, aimed at both professional ML developers and users without ML skills; it provides expert-prepared baseline models, allowing immediate work without writing code and additional configuration. In terms of security and privacy, it is officially stated to fully comply with ФЗ-152, use Cloud.ru encryption technology, and comply with standards for protecting confidential information. Cloud.ru describes itself as a leading provider of cloud and AI technologies, with 1400+ experts in IT, cybersecurity, and AI, providing proven methodologies and cloud migration best practices, as well as personal manager support.
Limitations
Regarding specific pricing, billing details (other than token-based billing), available regions, SLA, etc., official materials are not yet available.
