A controlled entry point
MyCove provides the desktop interface and a managed connection to approved backend services. Third-party browser apps should remain separated from credentials, native capabilities and administration tools.
For CIOs, IT leaders and community administrators
MyCove
AI & Agents
An installed access environment for tailored AI services built on Digital Village’s AI backend architecture. Connect users to authorized community information through an integration designed around identity, permissions and data separation.
Discuss an integration project →Security remains the foundation
Controlled access to useful community information.
MyCove’s primary role is resident protection within its installed desktop environment. A separately scoped integration can extend that environment with an AI assistant that answers questions from authorized handbooks, menus, schedules and service information, or with agents that perform explicitly permitted tasks.
Documents matched to the user
The backend verifies identity, community membership and document permissions before retrieving information. The AI receives only relevant material the requesting user is allowed to access.
Actions with defined authority
Read access does not grant permission to change records, share files or send messages. Agent tools need separate authorization, limited capabilities and confirmation for sensitive actions.
Architecture to define in the project
Identity first. Retrieval second. AI third.
- Authenticate the user. The installed client connects to the Digital Village backend using a verified identity. Define session expiry, stronger authentication for sensitive access and separate administrator roles.
- Enforce authorization on the server. Verify the user’s community and role on every request. Never rely solely on desktop restrictions or allow the AI model to decide access rights.
- Retrieve permitted document excerpts. Use retrieval-augmented generation (RAG) to search approved sources. Document permissions must apply to their searchable excerpts as well as the original files, and remain current as access changes.
- Generate an answer with sources. Send the question and relevant authorized excerpts to the selected model. Link answers to their source documents, with access checks when a user opens a source.
- Authorize any subsequent action. An agent’s request to use a tool must pass backend checks. Define permitted operations, approval requirements and an audit trail before enabling changes or external communication.
Access reflects the person’s role.
| User | Example information access |
|---|---|
| Resident | Resident-facing handbooks, dining information, schedules and community services. |
| Authorized family member | Shared community information and specifically authorized family or resident documents. Family status alone does not grant access to private information. |
| Staff | Operational documents permitted by the staff member’s role; restricted material remains inaccessible to other roles. |
| Another community | Its own authorized information, with no implicit access to the first community’s documents or conversations. |
Private models and private data need separate decisions.
A private model can help control where processing takes place. Local inference, community-hosted infrastructure and a privately operated backend each have different capacity and maintenance requirements. The project must specify the hosting arrangement and where document excerpts, conversations and logs are processed and stored.
Model hosting does not replace access control. Data separation must cover document storage, search indexes, embeddings, conversation history, caches and logs. Define retention, deletion, access revocation and whether any information may be used for model training. RAG allows source documents to be updated without treating them as model training material.
Documents and web content must be treated as untrusted input. Instructions embedded in retrieved material must not grant permissions or trigger agent actions. Backend authorization, limited tools and security testing are required alongside model safeguards.
A defined integration engagement
Agree the scope before connecting the data.
The scope of work should identify document sources and owners, eligible users, role permissions, supported use cases, model hosting, agent tools, data retention, audit requirements and operational responsibilities. It should also define deliverables and acceptance criteria for access enforcement, source accuracy and isolation between communities.
A focused starting point: a read-only assistant for approved community information, returning source-linked answers. Private resident documents and agents that change information can be scoped separately after the identity, retrieval and permission boundaries are established.
Define your community’s integration.
Bring your document sources, access requirements and intended use cases. We can discuss a separate project and scope of work built on Digital Village’s AI backend architecture.