_300x.jpg)
COVER FEATURE
MODEL CONTEXT PROTOCOL
MCP Servers: Capabilities and Risks
Having ceased to be just a trendy decoration that could be added to systems of any kind, AI is now an important factor of corporate strategy success. This has happened in part thanks to a unified protocol for communication of AI models with various IT systems.
Stanislav Makarov (s.makarov15@gmail.com), editor of Agentic Enterprise Telegram channel (Volgograd): Dmitry Volkov (vlk@keldysh.ru), senior fellow, M. V. Keldysh Institute of Applied Mathematics (Moscow).
An MCP Server as Easy as One-Two-Three
Using large language models that have no access to the current enterprise data and business processes creates a garbage in garbage out type of a situation. MCP servers have the potential to change this.
Alexey Arustamov (info@loginom.ru), Director, Loginom Company (Ryazan).
MCP: A New Era of Integration in the IT World?
An open protocol for LLM communication with external systems has standardized machine-to-machine interaction mediated by AI agents. However, MCP has certain deficiencies that hinder the protocol`s use in production systems.
Vladimir Bykov (vb@russianbi.ru), Director of Development, Gromov Circles Analytics and Research Center (Saint Petersburg).
MCP Gateway: A Unified Control Plane for AI Agents
A growing number of AI agents and MCP servers used in an enterprise calls for a centralized control hub. An MCP gateway can function as a unified interface for multi-agent systems.
Tatyana Sezemina (TSezemina@inno.tech), Head of AI Agent Platform, T1 AI Division, T1 Technology Group (Moscow).
DATA ECOSYSTEM
Unified Analytics Hub for City Transportation System
The Moscow transport creates huge amounts of data daily, from passenger traffic and road situation to telemetry systems, transportation infrastructure and city services. Critical factors include maximizing data processing speed, as well as ensuring the reliability of data and enabling its quick use to improve the efficiency of the city's transport system. The Unified BI Hub of Moscow Safe Transport Innovation Center facilitates the creation of the technology foundation for Russian capital development as a smart city.
Alina Malkovskaya (MalkovskayaAA@transport.mos.ru), Head of Interactive Analytics Division, Safe Transport Innovation Hub under the Center of the Traffic Organization, Moscow Transport Department.
AI AGENTS
A Quality Evaluation System for AI Agents
An AI agent does an employees job, and it should be held to the same standard. Putting someone on the payroll is not enough what matters is that they do the work well. An employee has a probation period, regular reviews and a manager who reads their correspondence with customers. An agent has one number instead: a share of good answers, obtained no one knows how. A company launches an agent, gets 76% good answers and cannot say who measured them or by what rules. This paper covers how agent quality assessment works in production: choosing between an in-house and a third-party model, designing criteria and labeling instructions, moving from team labeling to assessors, automating some criteria, tracking quality by topic, detecting degradation, and writing a knowledge base for an agent.
Alina Romanovskaya (denislina22@gmail.com), Discipline Head, a large fintech company (Moscow).
BUSINESS INTELLIGENCE
From Manual Flying to Real-Time Analytics
In a business environment where warehouse load, courier routes, and the status of thousands of orders change every minute, yesterday`s data is useless. A centralized analytics system offers the opportunity to move from manual flying mode to real-time management.
Dmitry Shirshakov (d.shirshakov@cdek.ru), Head of Big Data Department, CDEK (Moscow).
A Dashboard as a City Management Interface
With Moscow transportation system creating millions of digital events in real time, a key objective is achieving the high speed and accuracy of interpreting the continuous data stream. Analytics dashboards play a crucial role in addressing the challenge.
Timofey Penskoi (PenskoiTA@transport.mos.ru), Head of Design Directorate, Safe Transport Innovation Hub under the Center of the Traffic Organization, Moscow Transport Department.
SOFTWARE ENGINEERING
The Structural Risk of Open Source
Every one in four popular AI and data related open-source components gets through sudden licensing shifts, acquisitions, and deprecations. Is there a way to eliminate this systemic vulnerability built right into the foundation of modern enterprise AI stacks?
Dmitry Dmitrenko (dmitdm@gmail.com), independent expert (Moscow).
OPINION
From Data Collection to Consequence Awareness
Despite the colossal amounts of data accumulated by businesses and governments alike, crises still often come unexpected. Perhaps, the next stage of management development will involve not further data stockpiling, but rather gaining the ability to understand connections between data, risks, and their potential implications.
Mikhail Krikheli (info@itconsortium.ru), Chair of the Technology Companies of Russia Consortium (Moscow).
OS MEETING ROOM
From Broad Segmentation to Individual Engagement
When the number of marketing campaigns and channels grows faster than customers can comprehend, the winner would be not the one who sends the most messages, but the one with the ability to precisely target a specific customer. To improve efficiency and streamline customer value management workflows, Russian Lenta hypermarket chain has implemented a promotion optimization system enabling it to increase sales at less costs.
Nikolai Smirnov (nsmirnov@osp.ru), freelance writer (Moscow).
OS ACADEMY
The problem of comparing vector and relational DBMS
As semantic search systems grow in popularity, selecting the right DBMS for vector storage has become genuinely difficult: the list of candidates expands monthly, and each vendor's benchmark is designed to favor its own product. What should a practitioner choose a modern purpose-built vector DBMS or a time-tested relational DBMS with a specialized vector extension? This paper benchmarks eight vector and relational DBMS solutions under identical conditions within a unified RAG pipeline, using a corpus of regulatory and governance documents from HSE University as a real-world testbed.
Anastasia Milovanova (nastya.milowanowa2016@yandex.ru), student, MIEM HSE University; Egor Denisov (edenisov@hse.ru), lecturer, Faculty of Social Sciences, HSE University (Moscow).
