AI Actions

Integrate AI-powered automation into your workflows using Claude

Overview

Qontinui provides the AI Prompt action type that enables intelligent automation powered by Claude. Use it to add AI capabilities to your workflows for autonomous code analysis, fixes, and improvements.

When to Use AI Actions

Autonomous code analysis, fixes, and improvements
Test generation and documentation
Multi-step refactoring and migration tasks
Analyzing automation results and fixing issues

AI Prompt Action

AI_PROMPT

Execute AI Prompts with Claude

Execute an AI prompt with context isolation. The simplest and most powerful way to add AI capabilities to your workflow. Supports single-shot execution or auto-continuation for longer tasks.

promptstring

The prompt to send to Claude. Can be natural language or a slash command.

Example: Fix all TypeScript errors in src/components/

templateIdstring

Reference to a reusable prompt template (alternative to inline prompt).

Example: fix-type-errors

maxSessionsnumber | null

Maximum sessions to spawn. 1 = one-shot, null = unlimited auto-continuation until [TASK_COMPLETE].

Example: 1 (default, one-shot)

outputVariablestring

Store the AI output in a workflow variable for use by subsequent actions.

Example: ai_result

timeoutnumber

Execution timeout in milliseconds.

Example: 600000 (10 minutes, default)

Quick Bug Fix

Run a focused prompt to fix a specific issue

1.AI_PROMPT: 'Fix the null pointer exception in UserService.java'
2.SHELL: 'npm test' (verify fix)

Multi-Session Task

Allow AI to continue across sessions for larger tasks

1.Set maxSessions to null for unlimited auto-continuation
2.AI will spawn new sessions until it outputs [TASK_COMPLETE]
3.Results are accumulated across all sessions

Best Practices

Use one-shot for focused tasks

Set maxSessions to 1 (default) for quick, focused tasks that can complete in a single session. This prevents unexpected long-running operations.

Use auto-continuation for complex tasks

Set maxSessions to null for tasks that may need multiple sessions, like large refactoring or multi-file changes. The AI will continue until it outputs [TASK_COMPLETE].

Use output variables for chaining

Store AI output in workflow variables to pass information between actions. This is especially useful in sequences.

Combine with SHELL actions for verification

After AI makes changes, use SHELL actions to run tests, linting, or type checking to verify the changes are correct.