An AI agent promised in two weeks can take two months when integration work dominates delivery. For technology leaders, that means delayed business value and budget spent rebuilding connections across projects. ELEKS' Power Platform expert Sergii Kostenko explains how the Model Context Protocol (MCP) can reduce that repeated work in Microsoft Copilot Studio, and what changes for delivery, costs and governance.
Background & experience:
Sergii is a Power Platform expert with 15+ years of experience in enterprise architecture, Azure integration, SharePoint modernisation, ALM, CI/CD, and leading distributed teams.
1. Why do agent projects take longer than planned?
Sergii: Most Power Platform developers will recognise where the week goes, and it is rarely business logic. It is the plumbing: a system with no standard connector, so a custom one is built, described to the agent one action at a time, and rebuilt when the vendor changes an endpoint.
It is the main reason an "agent in two weeks" project quietly becomes a two-month one.
The Model Context Protocol (MCP) has been generally available in Copilot Studio since May 2025. My first reading of it: another way to plug in an API. That was wrong, and the people who approve agent projects should know why.
2. So what is MCP, and how is it different from a connector?
Sergii: A connector is a static list of actions, each selected, described to the orchestrator and wired by hand. With an MCP server, the AI agent asks the server what it can do at runtime and receives the current tools with descriptions. When the owner adds or changes a tool, the agent sees it during the next conversation.
That separates two jobs: owning the integration and building the agent. And with Microsoft shipping its own servers, Dataverse for Power Platform data and Work IQ MCP for Microsoft 365, much that was assembled by hand arrives as one "Add a tool" step.
3. How does MCP change a developer's day-to-day work?
Sergii: The change is where the hours go: less on Swagger files and authentication headers, more on what a client pays for, which is conversation design, guardrails and the edge cases where an agent must say "I can't do that" rather than guess.
Dataverse shows it well. Previously, list-rows, create-row and update-row actions were added and described separately, then chained through topic logic or a flow. Now the agent inspects a table's schema, queries it and creates a record in one conversation. In my own work, tool setup for a prototype request-intake agent fell from most of a day to under an hour. The tuning did not disappear; the plumbing did.
The same server is also reusable across compatible MCP clients such as VS Code and Claude.
4. What does MCP mean for the business side?
Sergii: For anyone responsible for a technology budget, three things follow.
- Time to value shrinks. Integration was always the hardest part of an agent project to estimate; with Microsoft-maintained servers, the first working version becomes a scope conversation, not an API one.
- Duplicated integration work goes down. One server can serve every agent in the company and compatible clients outside it. Someone still owns its API changes and support, but as one team's visible job, not a cost hidden in every project.
- Governance largely stays the same. In Copilot Studio, MCP servers run through the Power Platform connector framework, so existing data policies, authentication and telemetry apply. Work IQ and Agent 365-registered servers are governed from the Microsoft 365 admin centre instead.
5. What should teams do differently now?
Sergii: MCP in Copilot Studio is not another entry in the connector catalogue. It changes how integration work is shared between those who own systems and those who build agents and lets Copilot Studio agents join the wider agent ecosystem. The planning question is no longer "which connectors do we build" but "which systems already speak MCP, and which should."
FAQs
A connector is a wrapper around a system's API that gives an app, flow or agent a fixed set of actions, such as "get rows" or "send email". A developer selects, configures and describes each action by hand, and updates the connector if the API changes.
MCP (Model Context Protocol) is an open standard that lets an AI agent ask a system what it can do. The system's MCP server returns its current tools with descriptions, so nothing is wired by hand. If the owner adds or changes a tool, the agent picks it up in the next conversation.
In short, a connector is a static list of actions defined in advance, while MCP lets the agent discover them at runtime.
Not exactly, but they are closely related. MCP is the open standard. A Claude connector is how Claude uses it: most connectors (for Slack, Microsoft 365, Atlassian and similar services) are MCP servers that Claude connects to, so Claude can discover and use their tools without custom wiring.
So the word "connector" means something different here than in Power Platform. A Power Platform connector is a fixed, hand-configured list of API actions. A Claude connector plugs in an MCP server, as the article describes: the agent asks what the server can do at runtime.
In short, MCP is the protocol, and a Claude connector is the user-facing way to add an MCP server to Claude. Other compatible clients, such as VS Code or Copilot Studio can also use the same server.
Related insights
The breadth of knowledge and understanding that ELEKS has within its walls allows us to leverage that expertise to make superior deliverables for our customers. When you work with ELEKS, you are working with the top 1% of the aptitude and engineering excellence of the whole country.
Right from the start, we really liked ELEKS’ commitment and engagement. They came to us with their best people to try to understand our context, our business idea, and developed the first prototype with us. They were very professional and very customer oriented. I think, without ELEKS it probably would not have been possible to have such a successful product in such a short period of time.
ELEKS has been involved in the development of a number of our consumer-facing websites and mobile applications that allow our customers to easily track their shipments, get the information they need as well as stay in touch with us. We’ve appreciated the level of ELEKS’ expertise, responsiveness and attention to details.