MedaSystems Launches MCP Server in Beta, Bringing Expanded Access Program Data to AI Assistants
MedaSystems, the SaaS platform for managing Expanded Access, Post-Trial Access and Investigator-Initiated Study programs, today announced the beta release of the MedaSystems MCP Server. Using the server, authorized users can ask plain-language questions about their program data from the AI assistants they already use, such as Microsoft Copilot, Claude and ChatGPT, without switching tools.
The server is built on the Model Context Protocol (MCP), an open standard for connecting AI assistants to business systems. It inherits the same single sign-on, role-based permissions and data governance that customers rely on in MedaSystems today. If a user cannot see a record in MedaSystems, their AI assistant cannot see it either.
“Our customers already keep their program data in MedaSystems. The MCP Server lets them work with that data in the AI tools they are adopting, without exporting it or giving up the controls they depend on,” said Matt Rosen, CEO of MedaSystems. “We are releasing it as a beta so customers can shape it with us.”
Read More – From SEO to AI Visibility: The Next Evolution of Digital Discovery
What customers can do with it
Expanded Access programs generate a substantial body of records: requests, approvals, patient histories, biomarkers, dosing, safety reports, drug supply and outcomes. Bringing that information together has typically meant custom reports, spreadsheet exports and weeks of manual review. Within the MCP Server, a user can ask:
- “How many requests were approved, treated or withdrawn this year, by site?”
- “List each dose reduction and the treatment cycle in which it occurred.”
- “Summarize biomarker status for all treated patients and flag where it is unknown.”
- “Draft a case narrative for every patient treated for more than ten cycles.”
- “Which records are missing vital status or have inconsistent units?”
Beta capabilities include:
- Program and patient-level queries across cohorts, sites, investigators, treatment timelines and drug supply.
- Structured outputs, including listings, frequency tables and descriptive statistics, ready for medical and scientific review.
- Data-quality checks that surface missing values, duplicates, outliers and inconsistent terminology for follow-up by researchers or trial monitors.
- Scheduled reports, such as a weekly program status summary regenerated automatically from current up-to-date data.
- Support for every program type on the platform’s configurable data model: Expanded Access, Post-Trial Access and Investigator-Initiated Studies.
The server supports review and analysis by qualified professionals. It does not make clinical, regulatory or safety decisions.
Read More – SEO vs AEO vs GEO: The Future of Search Optimization for B2B Marketing in 2026
Built for regulated environments
- A dedicated connection for each customer. Each customer receives its own MCP server URL. MedaSystems works with the customer’s IT team to connect it to their approved AI assistant. Program data stays in MedaSystems, with no export to a separate system.
- Your permissions remain unchanged. Users authenticate with their existing MedaSystems credentials. Every request runs as that user, under the role-based access already configured in the platform.
- Read-only. The server retrieves data. It does not create or change records.
- Traceable answers. Each response identifies the source records it drew on and states whether results are complete or partial. Values that cannot be retrieved are flagged, not estimated.
- Privacy by design. Data minimization, access-controlled retrieval and no patient data in server logs.
“We built the server so the platform’s existing controls carry over unchanged,” said Jeff Thibeault, Chief Technology Officer of MedaSystems. “Users can see where every answer came from and whether it is complete. There is no separate, ungoverned path to sensitive data.”
Write to us [editorial@demandmediaagency.com] to learn more about our exclusive editorial packages and programmes.
