MCP for MSPs: What changes in service delivery

Richard McKay

Ask a technician what slows down a ticket and the answer is rarely the fix. It is finding the context: the alert in the RMM, the notes in documentation and the history in the PSA. 

MCP for MSPs means using Model Context Protocol (MCP) to connect an AI assistant to the tools an MSP runs on. The MSP sets the rules. The assistant asks the tools. The technician reviews the answer. The work shifts from navigating to deciding. 

This guide explains what changes in service delivery, what stays the same and how to adopt it without losing control of client data. Kaseya’s own roadmap points the same way: the next version of Kaseya Intelligence is described as an open platform built on APIs. 

What MCP for MSPs means 

MCP is an open protocol. An AI application, called the host, connects to one or more MCP servers, and each server exposes tools the model can call. The MCP architecture overview describes the host creating one client for each server. 

For an MSP, that maps onto the stack you already run. One server sits in front of the RMM. Another sits in front of the PSA. A third fronts documentation. The assistant connects to all of them and answers across them. 

This is already happening. Public registries list community-built MCP servers for Datto RMM, Autotask PSA, Kaseya VSA and Kaseya BMS. They are third-party projects, so the review is yours. 

One note on the term. Some vendors use MCP to mean something else, such as a management control plane. In this guide, MCP always means Model Context Protocol. 

Six places MCP changes service delivery 

Each of these is a workflow your technicians already run. MCP changes how they assemble context, not who is accountable for the result. 

Ticket triage with context 

A ticket arrives with a vague description. The assistant can pull the device’s open alerts from the RMM, check recent tickets for the same client and find the relevant documentation. Autotask PSA already connects tickets with endpoint data, documentation, passwords and remote access in one workflow. MCP extends that pattern to tools outside any single platform. 

Alert investigation 

Instead of opening the RMM and filtering by site, a technician asks what is failing at one client and why. The assistant queries the alerts, groups them and summarizes. The technician decides what matters. 

Documentation lookup and upkeep 

Documentation only helps if people can find it mid-ticket. An assistant connected to a documentation platform such as IT Glue can retrieve the right runbook or configuration note. It can also draft an update when something changes. A human approves the update. 

Remediation with approval 

Fixing is where the risk lives. An assistant can propose a script or job based on an alert. Running it should need technician approval, and the key behind the server should allow only what the task requires. Datto RMM ties write actions such as quick jobs to specific permissions, which gives that approval step something to enforce. 

Client reporting 

Quarterly business reviews eat hours of data gathering. An assistant can pull ticket volumes, patch status and backup health into a draft for the account manager to edit. The account manager still owns the story. 

Margin and utilization 

PSA data holds contracts, time entries and ticket volumes. A manager can ask which clients cost more to support than they pay. The answer is a starting point for a conversation. A human should check the numbers before they reach a client. 

What changes and what stays the same 

Adopting MCP shifts several things at once. Some of them are easy to miss. 

What changes: 

  • Technicians spend less time navigating and more time reviewing 
  • Integrations become assets to own, with keys, owners and review dates 
  • Governance becomes part of the service instead of a side project 
  • Tier 1 may handle questions that used to be escalated 

What stays the same: 

  • You remain accountable to the client for every action 
  • Contracts, SLAs and change processes still apply 
  • Consequential actions still need a human decision 
  • Documentation still has to be accurate before any assistant can use it well 

A four-step adoption ladder 

Most MSPs will do better climbing in steps than switching everything on at once. Each step adds capability and risk. 

  1. Ask. Connect one tool read-only and let technicians ask questions. 
  1. Draft. Let the assistant draft ticket notes, client replies and report sections for human review. 
  1. Act with approval. Allow narrow actions, such as running a known script, only after a technician approves each one. 
  1. Act within limits. Pre-approve a small set of low-risk, reversible actions with tight scope and full logging. 

Kaseya describes a similar logic for Kaseya Intelligence. In its Connect Europe 2026 announcement, agents identify issues, recommend remediation and, with technician approval, execute tasks. Early access is expected in late 2026, with general availability in 2027. 

The operating model: four questions 

Business of Tech’s December 2025 analysis argued that MSPs add the most value through discipline, not by building servers at random. Its questions still hold. Who owns the integration? What data flows through it? Where does authentication come from? Who shuts it down when it breaks? 

Multi-client work adds a fifth question: what keeps one client’s data away from another’s? An MSP’s RMM key often reaches every client. A prompt that mixes up two clients is a data-handling incident, not a typo. Name the client in every request, log every tool call and test with a deliberately confusing prompt before go-live. 

The NSA’s guidance on MCP warns that established defenses fall short of covering risks such as dynamic tool invocation and implicit trust. Treat each MCP server as a privileged integration. Use one key per integration, the narrowest permissions that work and logs you actually review. 

How to measure whether it works 

Pick a few numbers before the pilot starts: 

  • Time from ticket creation to first useful context 
  • Tier 1 escalations per week 
  • Ticket reopen rate 
  • Approvals requested versus approvals granted 
  • Technician hours reclaimed, tracked per client 

Baseline each one first. Without a baseline you are left with opinions. 

What this means for MSP revenue 

Kaseya’s 2026 State of the MSP Report found that 48% of MSPs rank AI and automation as their clients’ top need for 2026. Only 13% earn meaningful revenue from it so far. The same report found that 53% already use AI to automate ticketing, patching and monitoring. More than half have automated only about a quarter of their workload. 

The gap between demand and revenue is where service design happens. An MSP that can show scoped access, approval steps and tool-call logs can sell AI-assisted service with evidence behind it. One that cannot is selling a demo. 

Learn more about Kaseya’s MCP servers and skills.  

MCP for MSPs questions, answered 

What does MCP stand for in MSP? 

In this context, Model Context Protocol. It is an open standard that lets AI applications discover and call tools. Other vendors use the same letters for unrelated products, so check the context. 

Do MSPs need their own MCP servers? 

Not necessarily. Many MSPs will use servers that vendors or the community already publish. If you do build one, treat it as code you maintain, with an owner, a review date and a plan for updates. 

Will MCP replace my PSA or RMM? 

No. MCP is a way for an AI assistant to reach those tools. The PSA and RMM still hold the data and do the work. 

Is MCP safe for client data? 

Not by default. Safety depends on scoped credentials, approval steps and logging. Start read-only and add write access one action at a time. 

Where to start this week 

Pick one workflow, such as alert investigation for a single client site. Connect one tool read-only. Name an owner, set a review date and record your baseline numbers. Run it for a month before you add a second tool. 

Want to see how Kaseya connects service delivery, endpoints and documentation? Request a demo. 

Key Takeaways 

  • MCP connects AI assistants to your MSP stack under rules you set. 
  • Climb in steps: ask, draft, act with approval, then act within limits. 
  • Treat every MCP server as a privileged integration with an owner and a shutdown plan. 

 

Table of Contents 

One Complete Platform for IT & Security Management

Kaseya 365 is the all-in-one solution for managing, securing, and automating IT. With seamless integrations across critical IT functions, it simplifies operations, strengthens security, and boosts efficiency.

10 Dark Web Facts That You Need to Know

10 Dark Web Facts That You Need to Know

Read now
10 Facts About AI & Cybersecurity That You Need to Know

10 Facts About AI & Cybersecurity That You Need to Know

Read now
10 Facts About Phishing Risk & Dangerous Employee Behavior That You Must See

10 Facts About Phishing Risk & Dangerous Employee Behavior That You Must See

Read now