Skip to main content
In watsonx Orchestrate, agents are the core building blocks that execute tasks on behalf of users. An agent combines a large language model (LLM) with tools, collaborators, and optional knowledge sources to deliver intelligent, context-aware automation. Agents can be native, built directly in watsonx Orchestrate, or external, integrated from other platforms such as watsonx.ai, Salesforce AgentForce, or third-party frameworks via supported protocols.

​
What makes up an agent?

Every agent includes:
  • Name and Description
    The name uniquely identifies the agent, while the description serves two purposes:
    • It helps users understand the agent’s role and capabilities in the UI.
    • It informs supervisor agents when routing requests to collaborators. A well-crafted description should summarize the agent’s purpose, tools, and domain knowledge.
  • Instructions
    Instructions are natural language directives that shape the agent’s behavior. They define the agent’s persona, reasoning style, and how it should use tools and collaborators to solve tasks. Clear instructions ensure predictable, high-quality responses.
  • Tools and Collaborators
    Tools extend the agent’s functionality by enabling API calls, Python scripts, or other operations. Collaborators are other agents—native or external—that the agent can invoke to complete complex workflows.
  • Knowledge Base (optional)
    Agents can access domain-specific knowledge from uploaded documents or connected vector stores, improving accuracy for specialized queries.

​
Agent types

  • Native Agents: Built and managed within watsonx Orchestrate using the Agent Development Kit (ADK).
  • External Agents: Integrated from external platforms via protocols like external_chat or A2A.

​
What to do next

Descriptions and instructions

Learn more about agent descriptions and instructions, and how to write them effectively.

Authoring agents

Step-by-step guidance on creating your own agents, structuring their logic, and defining capabilities.

Connect to external agents

Explore how to integrate your system with external agents or services, and manage secure communication.

Importing and deploying agents

Learn how to import existing agents, configure deployment options, and publish agents to your environment.

Managing agents

Find out how to monitor, update, version, and control permissions for agents in your workspace.

Integrating agents with my application

Discover how to embed agents into your application using the embedded chat.