In November 2024, Anthropic released Model Context Protocol, or MCP, as an open-source standard for AI applications to connect to external systems. In December 2025, Anthropic transferred MCP governance to the Agentic AI Foundation, an industry group under the Linux Foundation. The governance transfer ensures no single company controls the protocol.
Since then, other major AI providers have adopted MCP, making it the connection standard. There are now more than 10,000 MCP connectors in use across finance, legal, software, and clinical research. It has become the standard way AI tools connect to business systems.
This post explains what MCP is, why it matters for clinical trials, and what sponsors and CROs should watch for as the protocol matures.
What is MCP?
Model Context Protocol, or MCP, allows AI models to connect to external systems and act on data inside them. MCP is an open-source standard that any vendor can adopt, giving developers a shared, standardized way to build connections between systems.
Since it’s an open standard that any vendor can adopt, MCP is becoming more common across clinical trial technologies.
Through a MCP connector, an AI tool can connect to:
- External systems and data sources, such as an EDC, RTSM, or CTMS
- Tools with calculations, such as randomization engines or supply forecasting calculators
- Workflows, such as pre-built prompts for drafting a site status update or a protocol deviation summary
How small sponsors and CROs can benefit from MCP
Before MCP, connecting an AI tool to a system such as an EDC or CTMS usually required building a one-off integration. Creating a new connection between systems can take a long time and require outside technical help, which can be a burden for small teams.
MCP connectors lower the barrier for technology adoption. Instead of a custom integration for every pair of systems, a vendor builds the connector once, to a common standard. Then any customer can turn it on without a separate integration project.
For small clinical trial teams, this means:
- Faster access to AI-assisted tools, without a lengthy IT buildout that a smaller budget can’t absorb
- Less dependency on technical resources to maintain integrations, which matters when there’s no dedicated data engineering staff to lean on
- More flexibility to add or swap AI tools without redoing integration work, which is useful when processes are still evolving
While MCP removes custom builds, there is still setup work required. This initial setup work allows you to decide who can use the MCP, what it’s allowed to see, and what it’s allowed to change.
How MCP can support clinical trials
MCP is still an emerging capability in clinical trials, so use cases are evolving. Other regulated industries, like finance and legal, have moved faster. For example, they are integrating AI tools into production systems for tasks such as fraud detection and case research.
Clinical trials haven’t quite reached that stage yet, but there are several possible use cases.
Possible use cases for MCP in clinical trials include:
- Plain language answers. An MCP-connected AI tool could query several systems, such as an EDC, an RTSM, a CTMS, and an eTMF. It could then answer a plain-language question like “Which sites are behind on enrollment and low on supply?”
- Faster status checks. A connector could surface information as a direct answer, updated from the source systems, instead of a manually assembled summary.
- Data-based documentation. Study documentation, like a summary of protocol deviations or a site status update, often starts as a blank page. An AI tool connected to the underlying data through MCP could draft a first version using the study’s actual numbers.
- Early detection of data inconsistencies. When the same information exists in more than one system, small mismatches can go unnoticed for a while. A connector with access to both systems could flag a mismatch as soon as it appears, rather than at the next manual reconciliation.
- Supply and inventory visibility. Supply managers often check inventory levels in one system and enrollment projections in another. An MCP connector could bring both into a single conversation with an AI tool, reducing back-and-forth between platforms.
Regulatory guidance around AI in clinical trials is evolving. Read our global regulatory AI article to learn more.
What’s next for MCP in clinical trials
MCP is a relatively new standard that opens exciting new possibilities for sponsors and CROs to accomplish more with less.
As with any technology in clinical trials, this protocol needs to be adopted responsibly. However, the learning curve is worth it since MCP could unlock meaningful new efficiencies in clinical research.
You can learn more about how to responsibly adopt AI in your clinical trials with the below resources.
- AI Key Questions Infographic: Learn what questions to ask vendors to safely adopt AI.
- State of AI in Clinical Trials eBook: Explore the full scope of the current state of AI in clinical trials.
Frequently asked questions about MCP in clinical trials
Find answers to common questions about MCP Connectors in clinical trials.
What does MCP stand for?
MCP stands for Model Context Protocol, an open standard for connecting AI models to external systems and data.
Is MCP specific to one company?
No. MCP is an open standard that any software vendor can adopt, similar to how many companies support a common file format or API standard.
How is this different from a typical API integration?
A standard API integration is usually built one connection at a time, between two specific systems. MCP aims to standardize that connection pattern so it can be reused across many tools and systems. It doesn’t replace your APIs — a connector usually sits on top of a system’s existing API and makes it available to any MCP-compatible tool.
Do MCP-connected AI tools work the same way as a chatbot?
No. A chatbot answers from general knowledge and doesn’t see your study data unless someone types it in first. MCP gives an AI tool a secure, permissioned connection to your actual systems, like an EDC or RTSM, so it can pull real, current data instead of relying on what a person types in.
Does an AI tool with a connector actually see my study data?
Yes. An AI assistant with no connection can only work from what someone types or pastes into it. With a connector, it has a permissioned connection to systems like your EDC or RTSM, so it works from real, current data.
Is this technology available in clinical trial software today?
Adoption is still early across the industry. Sponsors and CROs evaluating AI tools should ask vendors directly about their current MCP support and roadmap.
