MCP-Agents in Corporate Systems: How SimpleOne and Ainergy Integrated AI into Business Processes

Corporate AI-agents already know how to write emails, summarize meetings, and find answers in knowledge bases. But the SimpleOne team, together with technology partner Ainergy, went further: they embedded MCP-agents directly into the corporate platform itself — so that AI doesn't just advise alongside processes, but becomes their full participant.

The Boundary of Text Agents

A standard AI-agent in a corporate environment is a smart chatbot alongside the system, not inside it. It helps formulate tasks, summarize correspondence, or find the right documentation section. But creating a ticket in ITSM, checking incident status, finding a related object in the database, or reaching out to an external HR system — that's already beyond its capabilities.

As a result, the employee still switches between the agent and the corporate system, manually transferring data back and forth. AI helps you think, but doesn't help you act. For enterprise scenarios like ITSM, SDLC, or asset management, this is a fundamental limitation: what matters is not text, but concrete action in the system.

This very contradiction — an agent as consultant versus an agent as executor — became the starting point for SimpleOne and Ainergy.

MCP: Managed Access to Actions

Model Context Protocol (MCP) is an open standard developed by Anthropic that formalizes the interaction between a language model and external tools and systems. The key difference from direct API calls: the agent doesn't access the database directly. Instead, it receives a set of formally described tools with clear access rights and explicit execution logic.

Each tool is described as a contract: what it takes as input, what it returns, what rights are needed to call it. The LLM itself chooses the needed tool, calls it, and passes the result to the user — in a managed loop, without direct access to the system's internal architecture.

• retrieving data on a ticket, incident, or system object
• checking the current status of a task or change
• creating a new record — a ticket, task, related object
• finding related objects by specified criteria
• accessing an external system (ERP, HR, CMDB) with result returned to the agent

How It's Implemented in SimpleOne

In the SimpleOne platform, an MCP-server acts as a layer between the language model and business logic. Ainergy developed the GenAI platform infrastructure layer that allows describing MCP-tools at the level of SimpleOne's object model — without giving the LLM direct access to the database and internal platform logic. Practically, this means the same agent is configured differently for different roles.

An IT specialist sees tools for working with incidents and changes, a project manager — tools for tasks and statuses, an external user — only a limited set for submitting tickets. Rights to call tools are configured separately and tied to the user's role or operation context. The agent sees only permitted actions and works strictly within the boundaries set by the security team.

For an enterprise environment, this is crucial: auditability, security, and manageability are preserved, while AI gets real tools to participate in processes — not as a consultant, but as an executor with clearly limited rights.

What This Means

MCP architecture changes the role of AI in a corporate context: from a text assistant — to a participant in work processes. For platforms like ITSM, SDLC, and other enterprise systems, this opens a path to automating routine operations without compromising control over access rights and data security. Companies get AI that doesn't just think alongside the process, but acts inside it — within managed and auditable boundaries.