General
What are AI Agents?
Goal-driven AI Agents that operate autonomously within workflows. Fully customizable and adaptable, they can be tailored to support a wide range of use cases, across security, IT and more.
What distinguishes a deterministic workflow from an agentic workflow?
Traditional deterministic workflows require defining every condition and edge case, a rigid, time-consuming process. AI Agents bring adaptive intelligence that responds to change while maintaining boundaries and adhering to the rules you define.
Do AI Agents have guardrails?
Yes. Guardrails are explicit instructions you add to an Agent’s configuration to guide its behavior. They help prevent hallucinations, cut down on unnecessary verbosity, and keep outputs consistent and dependable for downstream automation. Agents always operate within defined boundaries: their assigned mission instructions and the toolbox of steps or workflows they are allowed to use. While they act autonomously, every action is logged in the Action Flow for full visibility and auditability. Guardrails are further reinforced by access policies and, when required, approval layers built into the tools the Agent can call.
What is the difference between AI Agents and Socrates?
AI Agents are fully configurable building blocks inside Torq workflows. They use AI models to take on a defined role, follow instructions, and call tools dynamically. They’re customizable and domain-specific, meaning you can tailor them to your exact SOC, IT, or operations use case.
Socrates, on the other hand, is Torq’s prebuilt virtual SOC analyst. It comes with opinionated defaults and ready-to-use skills, supporting Runbook planning, remediation, and case management.
Availability
Are AI Agents available on all Torq pricing tiers?
Yes. AI Agents are available across all tiers through a credit-based pricing model. This model combines tier-based entitlements with flexible add-on options, allowing teams to scale usage based on their needs. Whether you're just getting started or running large-scale workflows, you can tailor your AI Agent usage without moving to a different plan.
Functionality
How many tools can I connect to an AI Agent?
While there's no limit to the number of tools you can assign to an AI Agent’s toolbox, each execution is limited to a maximum of 15 tool calls.
Is there a limit to the number of AI Agents a customer can create?
Yes. Each workspace can create up to 50 AI Agents.
Can I count on consistent results from the AI Agent step when using the same instructions?
AI Agents are designed for adaptive, context-aware behavior, so their operation and outcomes may vary between workflow runs. If your use case demands identical results every time, consider using a more deterministic workflow design instead.