Skip to Content
Zeotap AgentOverview

Zeotap Agent

Zeotap Agent is Zeotap’s AI layer: a conversational agent that does CDP work for you, the same capabilities exposed as tools your own AI clients can call, and a set of guardrails deciding what any of it may do without a human saying yes.

Capabilities

FeatureWhat It DoesWho It’s For
AI AgentConversational interface that performs multi-step CDP operations on your behalfAll users — data engineers, analysts, operators
NL Audience BuilderDescribe an audience in plain English and get filter conditions you can apply into the builderMarketing analysts, growth operators
MCP ServerModel Context Protocol server exposing Zeotap as AI-accessible toolsDevelopers, AI engineers, automation builders
GuardrailsPolicies controlling what AI operations can do without human approvalPlatform admins, security teams

How It Works

Three layers sit behind everything above.

1. Tools

The MCP Server exposes Zeotap’s capabilities as structured tools. There are read tools — list models, get an audience, introspect a warehouse — and write tools that create audiences, trigger syncs, and configure destinations. Every call is authenticated and scoped exactly like the REST API.

2. The Agent

The agent chains those tools together to accomplish something you asked for in a sentence. Given “create an audience of users who spent over $500 in the last 90 days,” it reads your schema to find the right columns and computed attributes, builds the filter conditions, estimates how many members match, and hands the result back for you to apply — narrating each step as it goes.

3. Guardrails

Guardrails decide what the agent may do on its own. You write rules like “hold every deletion for approval” or “never let the agent touch these columns,” and they apply to every AI-initiated action regardless of which client made it — the chat inside Zeotap or an external MCP client. New workspaces start with three of them already switched on.

Getting Started

Chatting with the agent

  1. Click Zeotap Agent in the sidebar, or press ⌘J for the side panel
  2. Start a new chat and ask for something — “show me my largest audiences” is a fine first request
  3. Follow up in the same conversation to refine or extend what it did

Building an audience from a description

  1. Navigate to Audiences and click Create Audience
  2. Choose the parent model, then continue to the filter step — the agent panel opens by itself
  3. Describe the audience you want
  4. Review the suggested conditions and click Apply to builder
  5. Adjust anything by hand, then name and save

Connecting your own AI client

  1. Create an API key under Governance → AI API Keys in the sidebar, granting only the scopes your client needs
  2. Point your MCP client — Claude Desktop, Cursor, or anything else that speaks MCP — at your Zeotap MCP server
  3. Use Zeotap tools directly in your own AI conversations

See the MCP Server documentation for setup details.

Last updated on
OrganizationsAI Agent