Supported Targets — Unique AI Documentation

Supported Targets

1 min read

Supported Target Patterns

MCP Hub fronts any approved MCP-compatible origin server. What matters operationally is the set of tools the virtual server exposes, not the variety of underlying systems.

| Pattern | Examples | | Knowledge and collaboration | Confluence, SharePoint-like repositories, internal knowledge bases, document search, runbooks, and policy content. | | Communication | Outlook, email search, calendar lookup, draft-only communication workflows, and future human-in-the-loop send paths. | | Business systems | CRM, ticketing, service desk, customer context, operations workflows, and support knowledge retrieval. | | Data and market systems | Market data, research feeds, portfolio context, analytics APIs, and approved internal data services. |

Supported Client Patterns

| Client pattern | Example | Recommended endpoint strategy | | Unique-native | Unique Chat in Spaces | Use MCP Connectors and Space-level tool enablement. | | Enterprise copilots | Copilot, ChatGPT Enterprise, Gemini | Publish role-specific virtual servers with narrow read or draft capabilities. | | Developer tools | Cursor, IDE assistants, internal agents | Expose internal docs, tickets, and runbooks; avoid production mutation tools by default. | | Custom apps | Customer-built assistants | Integrate against the published MCP endpoint and hub OAuth metadata. |

Namespacing. When origins expose similar tool names, the hub presents clear namespaced tools so agents and users can distinguish the right action.