Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.cloud.cdata.com/llms.txt

Use this file to discover all available pages before exploring further.

​
The Three Zones

​
Zone 1: Your infrastructure (ISV-hosted)

This is the application and backend your team builds and operates. It includes:
  • Your AI experience or agent code
  • Your backend server, which generates and signs JWTs
  • The MCP client session that connects your agent to Connect AI Embed
  • Your LLM or AI model
  • Optional: application storage for persisting query results Your infrastructure is responsible for identity, access control, JWT lifecycle, user-level logging, and governing which of your end users can invoke AI features.

​
Zone 2: Connect AI Embed managed service (CData-operated)

This is the CData-operated platform your agent communicates with. It includes:
  • The MCP endpoint your agent calls
  • The User Management API, Metadata API, Query API, and Credentials Manager
  • The standard MCP tools available to your agent
  • Connection lifecycle management: authentication flows, token refresh, and credential storage CData operates this layer, including security, uptime, API maintenance, and connector updates. You configure it but do not host it.

​
Zone 3: Customer data sources

These are the systems your customers connect through Connect AI Embed. They include:
  • SaaS applications such as Salesforce, Jira, NetSuite, and HubSpot
  • On-premises databases and file systems (accessible via Connect Gateway)
  • Cloud data warehouses such as Snowflake, BigQuery, and Databricks
  • Hundreds of additional supported sources Data in Zone 3 is accessed live at query time. It is not copied, replicated, or persistently stored in the Connect AI Embed service.

​
How a Request Flows

A typical request from end user to data and back follows this sequence:
1
The end user inputs a natural language request in your application UI.
2
Your backend generates a signed JWT identifying the sub-account and scoping the request to specific connection IDs.
3
Your agent uses MCP tools to discover available sources and schema, then submits a structured query.
4
The Connect AI Embed MCP endpoint maps the tool call to the authenticated connection and executes the query against the live source.
5
Results are returned to your agent, which renders the answer for the end user. See Authentication in the API Reference for JWT configuration details. See MCP in the API Reference for the full tools reference and request flow example.