Action cards

  • Updated

Action cards make Optimizely Opal chat outputs interactive and actionable, so you complete end-to-end tasks, from analysis to execution, without switching tools or losing context.

Action cards replace plain text responses with structured UI elements that present information in scannable formats and support in-chat actions such as approving, editing, and submitting, all within the Opal conversation.

How action cards work

An action card is active and editable when:

  • It displays in the last Opal message in a thread.
  • No messages exist after that Opal message.

When you reply, the action card becomes read-only and preserves the interaction state at that point in the conversation.

Action card types

Action cards fall into four categories:

Standalone action cards

Standalone cards present common Opal outputs in structured formats. 

The following are a few examples of standalone action cards:

  • Form – Displays Opal-generated structured content, such as field-and-value pairs, in a reviewable layout. Use when Opal produces an output that maps naturally to named fields rather than prose.
  • Image – Displays AI-generated images inline with options to add to a task or download.
  • Table – Presents queried data in a scannable table with follow-up actions such as export, visualize, and create report.
  • Graph – Visualizes data with download and reporting actions.
Standalone table and graph cards display when a connected tool returns structured data. Conversational requests that do not invoke a data tool produce plain text or a code-generated image instead.

Optimizely product action cards

Product cards are rendered when a connected Optimizely tool returns a result and has a structured UI specification registered. The card replaces the plain text response with an interactive layout that lets you act on the output directly in chat, moving from insight to execution in one step.

Input action cards

Input cards collect required information before Opal runs a task, replacing multi-turn conversations with structured prompts.

  • Dynamic form – Presents all required inputs as structured form fields before Opal can execute a task. Differs from the standalone Form card in that Opal cannot proceed until the reader submits the fields.
  • Agent inputs – Presents tool-specific inputs, such as drop-down lists and text fields, when a tool requires them.
  • Image generation form – Collects the parameters required to generate an image.

Confirmation action cards

Confirmation cards present a Opal-drafted object for review and editing before Opal saves or executes it. Nothing is saved or sent until you click the confirm action on the card. When you reply after confirming, the card becomes read-only.

  • Create Skill card – Displays the skill Opal drafted, including its name, instructions, and activation conditions. Review and edit the fields, then click Save to create the skill without leaving the conversation. See Create a personal skill and Create an organization skill.
  • Create agent card – Displays the specialized agent Opal drafted, including its name, description, prompt template, and assigned tools. Review and edit the configuration, then click Create agent to register the agent.
  • Action confirmation card – Displays the details of an irreversible action, such as sending an email, before Opal executes it. Review the details and click Confirm or Send to proceed, or dismiss the card to cancel. Opal does not perform the action automatically.

Frequently asked questions

What is the difference between action cards and Canvas?

Canvas is a persistent workspace for longer-form content creation and editing. Action cards are inline UI elements within the chat thread, designed for structured outputs and single-action responses rather than extended editing sessions. The two features are complementary. An action card can include an action that opens or creates a Canvas. See Canvas overview.