Creating a Product
Products are the central hub for your AI features. A Product bundles your Flows and Agents into Capabilities, then exposes them to users through Surfaces — chat widgets, REST APIs, MCP servers, and more. Build once, deploy everywhere.
Before you begin
You’ll get the most value from Products if you already have at least one Flow or Agent built. If you’re starting from scratch, try the quickstart guide first, or create Flows and Agents directly from within your Product after you set it up.
Create a Product
Click Products in the sidebar.
Click Create Product.
Enter a name, such as Customer Support Assistant.
Add a description so your team can understand what the Product does at a glance.
Click Create.
You’ll land on the Product detail page, where you can start adding Capabilities and setting up Surfaces.
Choose a descriptive name that reflects what the Product does for your users, not how it works internally. This name appears in generated API docs and MCP manifests.
Quick start templates
Templates give you a fully configured Product with pre-built Capabilities and Surfaces, so you can go from idea to working deployment in minutes.
Click the Create dropdown on the Products page.
Select Quick Start.
Choose a template that fits your use case.
Follow the guided setup to customize it.
Templates are a good way to explore what Products can do before building one from scratch.
Adding Capabilities
Capabilities are the Flows and Agents that power your Product. From the Capabilities tab, you have several options:
New Flow — Build a Flow directly inside your Product.
New Agent — Create an Agent without leaving the Product editor.
Use Existing — Link a Flow or Agent you’ve already built.
Connect External — Connect an external A2A-compatible Agent, such as LangChain, CrewAI, or Vercel AI.
Each Capability gets a unique name and description that Surfaces use to present it to end users. For the full workflow, see Adding capabilities to a Product.
You can build and iterate on Flows and Agents right from the Product editor. You do not need to create them separately first. If you need to build one first, see Creating and editing Flows or Creating and configuring Agents.
Deploying with Surfaces
Surfaces are how your Capabilities reach users. Each Surface type is a different deployment channel:
Chat Widget — Embeddable chat interface for your website.
MCP Server — Expose Capabilities as MCP tools for AI assistants like Claude Desktop or Cursor.
REST API — HTTP endpoints with auto-generated OpenAPI docs for developer integrations.
A2A Agent — Agent-to-Agent protocol for AI-to-AI collaboration.
Email Channel — Send and receive emails to interact with your Capabilities.
You can add multiple Surfaces to a single Product and choose which Capabilities each Surface exposes. To learn how Surfaces work, see What are Surfaces?.
Start with a Chat Widget Surface to test your Capabilities interactively, then add an API Surface or MCP Surface when you are ready to integrate.
Product settings
From the Settings tab, you can configure:
Name and Description — Update how your Product is identified.
Icon — Choose a visual icon for quick recognition.
Status — Set to Active or Inactive.
Visualizing your Product
The Architecture tab gives you an interactive visual map of your Product. It shows how Capabilities connect to Surfaces, which makes it easier to understand and share your Product’s structure.
Next steps
Add Capabilities to your Product so users can interact with Flows and Agents through your Surfaces.
Set up a Chat Widget Surface to test your Product interactively.
Set up an API Surface if you want to give developers HTTP access.
Read What are Surfaces? if you are still deciding how you want to deploy.