Integrations & Data
An AI Employee (AIE) is only as useful as the systems it can reach. This section covers everything that connects an AI Employee to the world outside Ema: the Tools it calls to act in external systems, the data connectors that keep its Knowledge base current, the chat widget you embed on your own site, and the Data API that streams conversation data back into your analytics stack.
Everything here is configured from one place — the Integrations page in the AI Employee builder — and surfaced to AI Employees through two primitives:
- Tools — a single capability an AI Employee can invoke at runtime, such as "create a ServiceNow ticket" or "look up an order." Tools live inside integrations and are wired into a workflow through the agent's tool list. Throughout the platform this concept is always called a Tool — never an "action" or "external action."
- Knowledge base — synced documents an AI Employee searches to ground its answers. Data connectors populate Knowledge bases from sources like SharePoint, Google Drive, Confluence, and Box.
How the pieces fit together
| Layer | What it does | Where you configure it |
|---|---|---|
| Integrations Hub | Catalog of installable integrations, OAuth connections, MCP servers, and channels | Integrations page → All, MCP servers, Channels tabs |
| Tools | Individual callable capabilities inside an integration | Integration detail → Tools, then the Tool editor |
| Data connectors | Sync external documents into Knowledge bases | Integrations page → Data sources tab |
| Embeddable chat SDK | A branded chat widget for your own website | Integrations page → Channels tab |
| Data API | Export message-level conversation data to your warehouse | API key issued by an admin |
Under the hood, two services own these capabilities. The integrations service owns the catalog, installations, credentials, MCP server registration, and execution routing — it dispatches every Tool call to the right executor (http, mcp, composio, or script). The ingestion service owns data connector syncs, which run as long-lived background jobs rather than sub-second Tool calls. You never address these services directly; the builder UI and the agent runtime do it for you.
What's in this section
- Integrations Hub — the catalog, installing integrations, OAuth connections, MCP servers, and executor types.
- The Tool Editor — build a Tool from a schema plus an HTTP request or a TypeScript script, with input, credential, and setting references.
- Custom Integrations — wrap any external API as a custom integration and build its Tools end to end.
- Data Connectors — set up SharePoint, Google Drive, Confluence, and Box connectors to feed Knowledge bases.
- SharePoint Connector — sync a SharePoint Online site.
- Google Drive Connector — sync Google Drive via a service account or OAuth2.
- Confluence Connector — sync Confluence Cloud spaces and pages.
- Box Connector — sync Box files via OAuth2.
- Embeddable Chat SDK — embed a Conversational AI Employee as a chat widget on any site.
- Data API — stream message-level conversation data into your own BI stack.
Terminology. The capability an AI Employee invokes in an external system is a Tool. Older material may call it an "action," "external action," or "intelligent action" — those terms are retired. This section uses Tool throughout.