Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)
ReleaseAPI / Product · ActionableHabr AI→ original

Yandex 360 Opens 7 Office Services to AI Agents via MCP Server

Yandex 360 has launched an MCP server, a single "socket" through which AI agents from ChatGPT, Claude or Alice AI work with 7 office services, including Mail, Disk and Tracker. For example, an agent can find the latest email from a contractor and draft a reply, or build a project status report from a folder on Disk. The agent has the same permissions as the user, and it cannot send an email on a person's behalf. The ready-made server spares teams from building their own integrations.

AI-processed from Habr AI; edited by Hamidun News
Yandex 360 Opens 7 Office Services to AI Agents via MCP Server
Source: Habr AI. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceHabr AI ↗
Editor-in-Chief
Key metric #17 services are connected, from Mail and Disk to Wiki and Forms
Key metric #2Through MCP, agents can add employees and change the membership of working groups
Key metric #3Preview access is open to Yandex 360 customers and to paid plans

Almost every team now has an AI assistant, and almost every one of them hits the same wall: it knows nothing about the team's working data. Emails sit in the mail client, agreements in the calendar, tasks in the tracker, while the assistant sees only what someone pastes into the chat window. Yandex 360 is taking down that wall.

The company has launched an MCP server, a standard "socket" (MCP, Model Context Protocol, is an open protocol for connecting AI to third-party services) through which external AI agents get access to 7 services of the virtual office: Mail, Disk, Calendar, Telemost, Wiki, Tracker and Forms. In practice, you can tell an agent in ChatGPT or Claude, "find the latest email from the contractor and sketch a reply," and it will do so using real data rather than text pasted in by hand. For teams, this is a way to connect the office to AI without custom-built integrations.

What MCP is and what agents get

The easiest way to think of MCP is as a single socket. Instead of writing a separate integration for each neural network, a service exposes one MCP server, and clients connect to it on their own.

All seven services are available through this server. Yandex gives some clear examples of tasks. An agent can be asked to find the latest email from a contractor in Mail and prepare a draft reply. It can study the documents in a folder on Disk and compile a project status from them. It can create or reschedule a meeting in Calendar. It can read a project discussion in a chat and flag the open questions.

You can work with this both from external interfaces (ChatGPT, Claude, DeepSeek) and from Yandex's own products: Alice AI for business and agents built on the Yandex AI Studio platform. Developers can also connect Yandex 360 services to their own systems if they wish.

Where the access boundaries lie

The main question for any agent with access to email and documents is what damage it could do. Yandex has chosen a cautious design. Agents do not have unrestricted access to company data: they have the same permissions as the user, and they can use only the tools they are allowed to use. The press office offered a telling example: an agent cannot send an email on a person's behalf and can only compose a draft. The final click stays with the human.

There is also an administrative side. Through MCP, agents can manage an organization in Yandex 360 for business: for example, add new employees or change the membership of working groups. For HR and IT departments, this is potentially convenient routine automation, since a request for a new hire can be handed to an agent. But this is exactly where permission settings deserve the closest scrutiny.

It also matters that the MCP server runs on the Yandex 360 side. That means a team does not need to set up and maintain its own intermediary: it is enough to connect the ready-made server to the AI client of choice.

What this means in practice and what the limitations are

The server is currently in preview access. Any organization can connect, including Yandex 360 customers and users on paid plans. The company's announcement gives no details on request limits, pricing, or exactly how authorization is configured in each external client, so check the documentation before deploying.

The takeaway differs by role:

  • Developers: office data can now be connected to your own agents through a standard protocol, without inventing an integration from scratch.
  • Product managers: scenarios such as "compile a project status" or "prepare a reply to a counterparty" can be tested on the team's real data without waiting for separate development.
  • Executives and security teams: before giving an agent access, look at what permissions the user has. The agent inherits them in full, so broad employee permissions mean broad agent capabilities.

There is also a strategic signal. A Russian office suite is opening up to foreign AI interfaces, ChatGPT and Claude, while also offering its own Alice AI and Yandex AI Studio. For the market, this is another sign that office services are turning into a set of tools for agents, not just applications for people. Those building agent-based products for business should build this kind of integration into their plans now.

Story

Key Questions Answered

What MCP is and what agents get?

The easiest way to think of MCP is as a single socket. Instead of writing a separate integration for each neural network, a service exposes one MCP server, and clients connect to it on their own. All seven services are available through this server. Yandex gives some clear examples of tasks. An agent can be asked to find the latest email from a contractor in Mail and prepare a draft reply. It can study the documents in a folder on Disk and compile a project status from them.

Where the access boundaries lie?

The main question for any agent with access to email and documents is what damage it could do. Yandex has chosen a cautious design. Agents do not have unrestricted access to company data: they have the same permissions as the user, and they can use only the tools they are allowed to use. The press office offered a telling example: an agent cannot send an email on a person's behalf and can only compose a draft. The final click stays with the human. There is also an administrative side.

What this means in practice and what the limitations are?

The server is currently in preview access. Any organization can connect, including Yandex 360 customers and users on paid plans. The company's announcement gives no details on request limits, pricing, or exactly how authorization is configured in each external client, so check the documentation before deploying. The takeaway differs by role: Developers: office data can now be connected to your own agents through a standard protocol, without inventing an integration from scratch.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Hands-On AI Engineering · Hamidun Academy

Build & Ship Production AI Agents Without Boilerplate Code

Start your free 72-hour trial at Hamidun Academy (academy.hamidun.com) — interactive lessons on Claude Code & Cursor, live Tech Radar, and production agent templates.

Start 72h Free Trial →Vibe-Coding Intensive
What do you think?
Loading comments…