A connection to tools, with a person behind it
Model Context Protocol, or MCP, is a way for compatible AI software to discover and use tools provided by another system. In a practice system, those tools might find an authorised participant record, prepare a case note or create a follow-up task. What matters is which tools are exposed and whose permissions apply.
Ask four access questions
- Which employee and organisation does this connection represent?
- Can it only see the participants that employee can access?
- Which actions can it take, including financial or record changes?
- How can the organisation revoke it and inspect what it did?
A connection should not become a second, broader route into participant information. Test with a restricted staff role, not only an administrator.
How CordoCare approaches the connection
An owner or administrator enables organisation MCP access. Employees create personal, expiring keys, and CordoCare applies their current membership, role and participant scope to requests. Supported writes have audit records and repeat-safe request controls. Hard deletes and owner-role changes are unavailable through these agent tools.
Start with a bounded job
Try a synthetic record and one task, such as preparing a follow-up. Inspect the result and audit history, then revoke the connection and confirm it stops working. Separately assess the AI client’s privacy and data handling before using participant information.
See CordoCare’s agent access and controls or request a practical demonstration. MCP describes a connection standard; it does not by itself establish privacy compliance or safe automation.