Working with Agents

Agents are the core building blocks of Aethyr. Learn how to create, configure, and orchestrate AI agents for your workflows.

Creating Agents

Agents can be created through the Console UI or programmatically via the API. Each agent has a unique configuration that determines its behavior.

Via Console UI

  1. Navigate to Agents in the sidebar
  2. Click Create Agent
  3. Fill in the agent details (name, model, system prompt)
  4. Configure tools and capabilities
  5. Save the agent

Programmatic Management

Agents can also be created and managed programmatically. Programmatic access is scoped and provisioned per deployment—your team will walk through the details during onboarding.

Agent Configuration

PropertyTypeDescription
namestringDisplay name for the agent
modelstringModel ID (e.g., claude-sonnet-4, gemini-2.5-pro, or a local model)
systemPromptstringInstructions that define agent behavior
temperaturenumberCreativity level (0-2, default 0.7)
toolsstring[]MCP tools the agent can use
maxTokensnumberMaximum response length
knowledgeBasestringRAG corpus ID for context retrieval

Multi-Agent Workflows

Aethyr supports multi-agent orchestration where multiple specialized agents collaborate on complex tasks. Agents can be chained, run in parallel, or coordinate through a supervisor agent.

Sequential

Agents execute in order, each receiving the output of the previous agent.

Parallel

Multiple agents work simultaneously on different aspects of a task.

Supervisor

A coordinator agent delegates tasks to specialized sub-agents.

How Orchestration Works

A supervisor agent plans a task and delegates sub-tasks to specialized agents, each with its own model, instructions, and scoped set of tools. Sub-agents run their own reasoning loops and return results to the supervisor, which synthesizes the final response. Delegation depth and tool access are governed per agent.