OpenAI agent builder + Tinybird MCP: a data-driven agent workflow

OpenAI has introduced AgentKit, a comprehensive toolset for developers and enterprises to create, deploy, and optimize AI agents. This tool is complemented by OpenAI's Agent Builder, a UI-based workflow system that integrates OpenAI models with various tool calls, enabling cross-functional teams to design agent chains with conditional logic and connectors like the Tinybird MCP Server. The Tinybird MCP Server hosts analytical tools that allow for resource exploration and query execution against Tinybird data sources, enhancing the capability of agents to provide deterministic, data-driven responses. The workflow involves categorizing user inputs into predefined categories—such as "sales_performance" or "inventory"—and routing them to specialized agents with fine-grained access to relevant data, ensuring efficient and accurate responses. By leveraging token-based authorization and a structured output schema, the system facilitates robust data exploration and response generation. This setup is particularly useful for incorporating chat-based data exploration into applications like an e-commerce revenue dashboard, moving beyond traditional dashboards to more dynamic, chat-driven analytics.