Skip to main content
Agents perform intelligent work within your organization. In Thygon, an agent is an autonomous reasoning unit that executes cognitive tasks, makes decisions, and coordinates tool interactions. Rather than existing as disconnected chat interfaces, Thygon agents are embedded directly into your workflows to automate intelligent processes.

​
Role in the organization

Agents represent the “brain power” of your automated operations. You can assign agents to:
  • Triage and route information: Analyze incoming events or documents and classify them (e.g., support ticket triage).
  • Make decisions: Determine the next operational steps based on context and organizational policies.
  • Coordinate tools: Interact with external systems through Connections to read or write data.
  • Collaborate with humans: Draft responses, summarize issues, and request human feedback or approvals.

​
Configuring agent nodes

Agents are defined as specific nodes within your workflow graph. Each agent node configuration includes:
  • Identity: A unique identifier and display name for auditing.
  • System instructions: The prompt or system instructions that guide the agent’s behavior.
  • Connections: The external systems the agent is allowed to access and interact with.
Here is an example of an agent node in a workflow definition: 
name: triage_ticket
type: agent
agentId: support-classifier
instructions: |
  Analyze the support ticket payload.
  Classify the priority as low, medium, or high.
  Identify the key product area associated with the ticket.
connections:
  - slack
  - linear

​
Auditable execution

Every decision, tool invocation, and event published by an agent is recorded in the immutable event log. This provides full traceability, allowing you to audit how an agent arrived at a decision and verify its compliance with organizational guidelines.