Skip to main content

Agents

The Agents page is the central hub for building, configuring, and managing enterprise AI agents in Aegis.

An Agent brings together System Instructions, a Model Alias, Runtime Bindings, and Knowledge Base (RAG) connections into an executable AI service.

Agent Workspace Overview​

The Agent Workspace toolbar provides 5 specialized management sections:

  • Agents: Configure and deploy active runtime agents.
  • Prompt Library: Reusable global prompts maintained by IT.
  • Runtimes: Container execution environments and discovery contracts.
  • Tools: Approved enterprise tools catalog and credentials.
  • Knowledge Base: External document stores and vector indices.

Agent Configuration Parameters​

When creating or editing an Agent, configure the following properties:

  • Name (Slug): Unique identifier for the agent (e.g., hr-assistant).
  • Description: Summary describing the agent's responsibilities.
  • Target Runtime: Select from active registered Runtimes.
  • Runtime Agent ID: Target agent identifier on the runtime container (e.g., hr_assistant).
  • Agent Type:
    • Atomic (atomic): Single-step task agent.
    • Workflow / Composite: Multi-step orchestration agent.
  • Model Alias: Target model alias registered in AI Gateway (e.g., gemini-flash).
  • System Instruction: Core prompt instructions governing agent behavior.
  • Knowledge Bases: Select team Knowledge Base RAG datastores to attach.

Creating a Manual Agent​

  1. Navigate to Agent Workspace > Agents in the Aegis Console.
  2. Click + Create Agent (navigates to the Manual Agent Builder).
  3. Fill in Name, Description, and select a Target Runtime.
  4. Select a Model Alias and enter the System Instruction.
  5. Select target Knowledge Bases to attach to the agent.
  6. Click Save Agent.