Actions3
- Agent Actions
- AI Interaction Actions
AI Interaction → Invoke
AI-generatedSummary
Invokes a general-purpose AI interaction via the Port API, optionally managing approval workflows.
Inputs
- userPrompt (required) — The text prompt to send to the AI interaction.
- tools (required) — Array of tool names as a JSON string (supports regex patterns).
- executionMode — Optional execution mode: 'Automatic' or 'Approval Required'.
- invocation_identifier — Optional invocation identifier for tracking.
- labels — Optional labels object as JSON.
- model — Optional AI model: 'gpt-5', 'claude-sonnet-4-20250514', or 'claude-haiku-4-5-20251001'.
- provider — Optional provider: 'anthropic', 'azure-openai', 'bedrock', 'openai', or 'port'.
- systemPrompt — Optional system prompt to guide the AI's behavior.
Output shape
An object containing invocationIdentifier, events, executionMessages, finalData, executionMessage, and data.
The response is parsed from Server-Sent Events (SSE). If the Tools field is not a valid JSON array, an error is thrown.
Examples
Example 1: Invoke an AI interaction with approval required
User Prompt: 'Analyze the user permissions', Tools: ['^(list|get|search|track|describe|count|run_*)_.*'], Execution Mode: 'Approval Required'
Example 2: Invoke an AI interaction using a regex for tool matching
User Prompt: 'Summarize recent tasks', Tools: '["^(list|get|search|track|describe|count|run_*)_.*"]', Model: 'claude-sonnet-4-20250514', Provider: 'anthropic'