What are Products?
Products are organizational containers that group your AI capabilities (Flows and Agents) and make them available through deployment channels called surfaces.
Why Products exist
Products solve a key problem: you've built Flows and Agents, but how do you expose them to users? Products connect your AI capabilities to real-world interfaces like chat widgets, REST APIs, and MCP servers.
Instead of deploying each Flow separately, you create a Product that bundles related capabilities and deploys them together through one or more Surfaces.
Product structure
Every Product has three components:
Capabilities — The Flows and Agents you want to make available
Surfaces — How users access those capabilities (chat, API, MCP, A2A)
Configuration — Settings like environment, status, and versioning
Example: Customer Support Product
Maya runs support for a SaaS company handling 200+ tickets daily. Her team spends hours answering repetitive questions while urgent issues wait in the queue. She needs to automate the simple stuff without losing the human touch on complex problems.
She starts by identifying what to automate. About 60% of incoming questions are FAQ-style—password resets, billing questions, feature availability. Another chunk needs immediate routing based on urgency and topic. She decides to build a customer support Product with three capabilities:
FAQ answering Flow — Handles common questions instantly using her documentation as context
Ticket routing Agent — Analyzes incoming requests and assigns them to the right team member based on topic and sentiment
Documentation search — Retrieves relevant help articles when questions fall outside the FAQ scope
Next, she thinks about where customers actually ask for help. Most use the website chat widget, but her mobile app also needs support, and her team works primarily in Slack. Rather than deploying each capability separately across three places, she creates one Product with three Surfaces:
Chat widget — Embedded on the website for instant FAQ answers
REST API — Powers the mobile app's support interface
Slack integration — Routes tickets and surfaces answers directly in her team's workspace
She deploys it to production, sets the status to active, and tracks it as version 1.0. Within the first week, she notices the FAQ Flow is handling 40% of inquiries end-to-end, and the routing Agent is categorizing tickets with 85% accuracy. Her team's response time drops from 4 hours to 45 minutes.
Two months later, she adds sentiment analysis to flag frustrated customers for immediate escalation—same Product, now version 1.2, with an upgraded routing capability.
A single Flow can be used in multiple Products. For example, your "summarize text" Flow might be part of your customer support Product, your content creation Product, and your research assistant Product.