Create Agent¶
Creates and returns a single agent configuration that defines the model, available tools, role, and system prompt. This configuration is intended to be passed into downstream nodes that run one or more agents.
Usage¶
Use this node to define an agent before orchestrating multi-agent interactions. Typically, you will create one or more agents with different roles and prompts, then connect their outputs to a team/execution node (e.g., Run Agents Team) to perform collaborative tasks.
Inputs¶
| Field | Required | Type | Description | Example |
|---|---|---|---|---|
| model | True | CHOICE | Select the LLM model the agent will use for inference. The available options are provided by the platform and may vary over time. | gpt-4o |
| tools | True | MULTICOMBO | Choose external tools the agent can access during execution. Enable one or more tools as needed for the agent's tasks. | {'Calculator': True, 'WebSearcher': True, 'WebScraper': False, 'DateTime': True} |
| role | True | STRING | A short description of the agent’s role. This is prepended to the system prompt and can help orchestrators select the right agent for a given task. | A meticulous research analyst specializing in market trends and competitor intelligence. |
| system_prompt | True | STRING | The detailed system prompt guiding the agent’s behavior, tone, knowledge scope, and rules. | You are a helpful AI assistant. Provide concise, well-cited answers. When uncertain, ask clarifying questions before proceeding. |
Outputs¶
| Field | Type | Description | Example |
|---|---|---|---|
| agent | AGENT | The constructed agent configuration containing the selected model, tools, role, and system prompt. Connect this to nodes that run agents or assemble agent teams. | {'model': 'gpt-4o', 'tools': {'Calculator': True, 'DateTime': True}, 'role': 'A helpful AI assistant.', 'system_prompt': 'You are a helpful AI assistant.'} |
Important Notes¶
- Role is prepended to the system prompt and may be used by orchestration logic to select or prioritize agents.
- Tools are optional per capability—enable only those necessary for the agent’s tasks to keep behavior focused.
- Model and tool options are provided by the platform and can change; verify selections if a workflow is reused later.
Troubleshooting¶
- Selected model is missing: Reopen the model dropdown and choose an available model. Options may have changed since the workflow was created.
- Tools not taking effect: Ensure the specific tools are enabled (set to true) in the MULTICOMBO input and that the downstream node supports tool usage.
- Downstream node rejects the agent: Confirm the output from Create Agent is connected to an input expecting an AGENT type and that all required fields are set.