Limited Release
This action is in limited release and is not generally available.
Invoke AI Agent
The Invoke AI Agent action runs a pre-configured AI agent that can reason about tasks and call custom workflows as tools. This is the most powerful AI action, enabling intelligent automation that adapts to context.
When to Use This Action
Use this action when:
- You need the AI to perform multi-step reasoning: querying data, making decisions, and taking actions based on the results
- Your task requires calling one or more custom workflows as tools (e.g., looking up CRM data, creating records)
- You want an AI agent that can handle complex, open-ended instructions autonomously
Don't use this action for:
- Simple text analysis or classification: Use Prompt AI instead, which is simpler and faster for single-turn AI tasks.
Configuration
| Field | Description |
|---|---|
| AI Agent | Select the AI agent to run. Agents are configured in your MSPintegrations settings with specific capabilities, tools (custom workflows), and behavioral instructions. |
| Prompt | The prompt or instruction to send to the AI agent. Supports text expressions with {{variable}} syntax. Be specific about what you want the agent to accomplish; the agent decides which tools to use. |
| Conversation ID | A unique identifier to maintain context across multiple interactions. Use this when you want the agent to remember previous exchanges. Leave empty for single-turn conversations. Example: telegram:{{event.message.chat.id}} |
| Expanded Response | When enabled, wraps the response in an object: {"Response": "<the agent's final response>"}. Useful when you need to access the response by name (e.g., {{custom.result.Response}}). |
How Agents Work
- The agent receives your prompt along with the available workflow data
- It analyzes the request and decides what actions to take
- It calls custom workflows you have configured as tools
- It continues reasoning and calling tools until the task is complete
- It returns a final response
Returned Variables
When you configure Store the results in Variable:
| Variable | Type | Description |
|---|---|---|
| (root) | string or object | The agent's final response. Format depends on the agent's configuration and output schema. |
If the agent is configured with a structured output schema, the response will be a JSON object matching that schema.
Example Use Cases
Intelligent Ticket Routing
An agent configured with tools to:
- Query the CRM for customer information
- Check ticket history
- Assign tickets to appropriate queues based on context
Automated Response Generation
An agent that can:
- Search a knowledge base
- Query ticket history
- Generate contextual email responses
Multi-Step Data Processing
An agent that orchestrates:
- API queries across multiple systems
- Data transformation and comparison
- Conditional logic based on results
Constraints
| Limit | Value |
|---|---|
| Maximum agent turns (tool calls) per invocation | 100 |
| Maximum agent nesting levels | 10 |
| Conversation history | Automatically summarized to stay within processing limits |
Notes
- No external AI account or API key is required. AI actions are provided by MSPintegrations.
- The agent nesting limit prevents infinite recursion between agents