Model Context Protocol for AEC: Part 1

Introduction

AI and large language models (LLMs) are beginning to be identified as the fourth industrial revolution. Their impact is felt across many industries, with construction being no exception. The year 2025 has been notably defined as the year of AI agents. AI agents are software systems using AI to pursue goals and complete tasks on behalf of users.

At the end of 2024, Anthropic released the Model Context Protocol (MCP), a protocol designed to easily allow foundational AI models (LLMs) to use external tools to expand their context. We won’t spend time here explaining what MCP is technically, as many comprehensive articles already exist. For a deeper dive into MCP, refer to this link. For a simple description, MCP has often been described as

“USB-C for AI integrations,” providing a universal way for AI models to connect to different devices and data sources.

Current landscape of MCP

If you’re following recent developments around AI agents, MCP represents a significant new wave addressing broader issues beyond the traditional “island of agents” problem.
Today’s agents often operate in isolated environments, lacking effective communication or the ability to share valuable information, resulting in fragmented tool usage and inconsistent frameworks. MCP significantly enhances agent intelligence by providing standardised access to context, simplifying how agents invoke APIs, call functions, and integrate with external systems—essentially empowering them to better think, act, and interact in their environment. Since its release, we’ve seen many big players quickly adopting MCP, along with ongoing debates about whether MCP is the definitive solution. While MCP may not ultimately become the final standard, at this point in time, it is a powerful solution fueled by a vibrant open community of strong developers from various major companies. These positive factors encouraged us to explore MCP further.

MCP’s strategic value in AEC

The AEC industry often struggles with efficient and effective API consumption, primarily due to knowledge gaps around integration and practical usage. MCP addresses these challenges by simplifying interactions, allowing LLMs to consume the correct tools for users effortlessly. Essentially, these tools serve as wrappers around API functions. Although MCP is frequently compared to APIs, the two are distinct, even though they share similarities.

MCP serves as a context management layer that enables LLMs to interact more effectively with external systems. In contrast, APIs provide secure and reliable connections to these systems.

MCP doesn’t replace APIs; rather, it provides a standardised layer on top, often wrapping around existing APIs to facilitate easier and more dynamic interactions.

As AI technologies evolve, the potential benefits for streamlined workflows, enhanced decision-making processes, and overall operational efficiencies become increasingly evident. These improvements offer compelling reasons for organisations to follow developments in MCP and related interoperability standards closely.

• Automation of routine tasks: MCP enables teams to automate routine construction management tasks, allowing personnel to focus on more strategic activities.
• Unified tools: MCP’s interoperability simplifies the integration of various applications, reducing the complexity traditionally associated with multiple software integrations.
• Enhanced decision-making: Access to comprehensive data and insights provided through MCP facilitates quicker and more informed decision-making. For example, project managers could have instant access to critical project statistics via an AI assistant, enabling strategic decisions that positively impact project timelines and budgets.

Practical application: Autodesk Platform Services (APS)

Recognising MCP’s strategic benefits, we saw significant value in creating an MCP solution specifically for Autodesk Platform Services (APS). APS is a robust solution, offering APIs to operate and query project resources and data.

In the AEC industry, it’s critical for project managers to access accurate information in the simplest possible way. With the rise of LLMs, the method of accessing this information is shifting toward a more conversational approach. By equipping an LLM with specialised APS tools, a project manager can easily extract essential information through simple conversational prompts. For instance, a few messages could generate a chart detailing project deliveries and installations scheduled for the month, visualise the latest model updates, or identify clashes between models.

Leveraging APS with MCP, teams can expect simplified access to critical project data from multiple sources. Imagine an AI assistant capable of retrieving real-time project schedules, resource allocations, and bid analyses without the need to switch between various software platforms. MCP can facilitate a collaborative environment where team members benefit from AI-driven insights across APS services and other applications. Such interoperability could significantly streamline decision-making and enhance team alignment by presenting relevant information cohesively.

Additionally, an extra benefit is that LLMs handle unstructured data effectively. Even if a model lacks perfect data integrity, LLMs can interpret, clean, structure, and better utilise data based on its semantic meaning. MCP allows different applications to connect with LLM clients, enhancing their capabilities and facilitating chained operations. It enables workflows across services in a way that previously required entire custom applications. MCP could significantly accelerate the industry’s shift to a fileless environment, promoting smoother and more fluid data exchange and enabling micro-extraction and exchange of data. Under an MCP framework, AI assistants could provide context-sensitive recommendations based on project statuses tracked within APS, helping teams anticipate challenges and seize opportunities more effectively. For example, an AI-powered tool could seamlessly update information within documents or models based on user inputs or ongoing project changes. Because MCP standardises how agents discover and call tools, it also makes it trivial for a single generalist agent—or a coordinated team of domain‑specific agents—to query a BIM model, pull quantities, schedules, or clash data, enrich that information with live feeds from IoT sensors or cost databases, and push the results straight into any dashboard. In effect, MCP acts as the connective tissue that allows many services to cooperate over a common data layer, accelerating the fusion of external datasets, BIM, IoT and AI.