Connect AI documentation · Full API reference & configuration details

Your first answer in 5 minutes

Connect a GitHub account to Connect AI in one click, wire the MCP server into Claude Code, and ask Claude Code a natural-language question about your repos.

Using Cursor, Claude Desktop, or a different MCP client instead? Skip to the client setup matrix in the MCP reference.

What you need

• Node.js 18+ (Claude Code is distributed as an npm package)
• A Connect AI Developer Edition account (free, no credit card)
• A GitHub account (any tier)

---

Step 1: Connect GitHub via Data Copilot

Sign in to cloud.cdata.com. The default landing is Data Copilot, which has three pre-wired quick prompts: GitHub, Slack, and Google Calendar.

1. Click the GitHub quick prompt
2. Connect AI opens the GitHub OAuth flow in a new tab. Sign in and authorize.
3. The connection is created automatically and Data Copilot runs an example query against it

You can use Slack or Google Calendar instead. Same flow. Pick whichever has the most data for you to ask about.

Step 2: Create a Personal Access Token

1. Click the gear icon (top-right) → Access Tokens
2. Click Create PAT, give it a name, and copy the token immediately, it’s only shown once

Step 3: Base64-encode your credentials

The MCP server uses Basic auth. Encode email:PAT locally:

echo -n "[email protected]:YOUR_PAT" | base64

On Windows (PowerShell):

[Convert]::ToBase64String([Text.Encoding]::UTF8.GetBytes("[email protected]:YOUR_PAT"))

Copy the resulting string.

Step 4: Install Claude Code and register the MCP server

npm install -g @anthropic-ai/claude-code

Register the Connect AI MCP server:

claude mcp add --transport http connectmcp https://mcp.cloud.cdata.com/mcp \
  --header "Authorization: Basic YOUR_BASE64_STRING_HERE"

Verify it’s registered:

claude mcp list

Step 5: Ask your first question

Start Claude Code in any project directory:

claude

Ask:

Claude Code calls getCatalogs and lists GitHubConnection: the catalog name the Data Copilot quick prompt creates. Now ask something with real data:

Claude Code calls getSchemas and getTables to discover the GitHub data model, generates SQL, runs queryData, and returns a plain-English answer. The tool calls are visible inline; Claude asks for confirmation by default before executing each one.

What you can ask

With GitHubConnection connected, useful prompts include:

The same pattern works with any source you connect: Salesforce, HubSpot, Workday, NetSuite, Snowflake, Postgres, and hundreds of others. Each Dev Edition account supports up to 5 connected sources.

What’s next

• 10-minute scaffold: generate a working agent project (LangChain, Google ADK, or Vercel AI SDK) wired to Connect AI
• MCP reference: full tool list, plus client setup for Cursor, Claude Desktop, n8n, and more
• By source: source-specific quirks and example queries for Salesforce, HubSpot, Workday, and NetSuite