GrowthLoop MCP
Headless usage of GrowthLoop with MCP
The GrowthLoop MCP server lets AI assistants securely work with your GrowthLoop account through the Model Context Protocol (MCP).
With MCP, an assistant such as Claude/ChatGPT can discover GrowthLoop tools, call approved GrowthLoop API endpoints, and use Marve for agentic audience and data questions.
What You Can Do
Use the GrowthLoop MCP server to:
- List and inspect dataset groups, audiences, journeys, destinations, and exports.
- Create and update supported GrowthLoop resources through public API tools.
- Ask Marve questions about audience strategy, available warehouse data, dataset groups, and audience-building ideas.
- Continue Marve conversations with a
chat_idwhen your MCP client supports follow-up context.
MCP Tools vs. Marve
GrowthLoop exposes two kinds of MCP capabilities:
| Capability | Use it for |
|---|---|
| Public API tools | Direct actions such as listing audiences, creating audiences, viewing exports, or updating journeys. |
| Marve agent | Open-ended, agentic questions about data, audiences, segmentation strategy, and campaign ideas. |
Use direct API tools when you know the exact GrowthLoop action you want to take. Use Marve when you want an AI agent to reason through the request.
Security
The MCP server uses your GrowthLoop API credentials. The assistant can only access resources available to that API key and organization.
Marve uses GrowthLoop metadata and warehouse schemas to reason about audiences and data. GrowthLoop does not copy row-level warehouse data into the application database.
Updated 25 days ago