Type: nodetool.agents.AgentStep
Namespace: nodetool.agents
Description
LLM step that inherits the workflow’s configured model. agents, llm, step, reasoning
Use this node inside agent-mode workflows when you need an LLM to
reason, transform text, or call tools. The step receives upstream
edges as context, runs the workflow's configured LLM, and emits the
final text on its `output` handle.
Properties:
- instructions (required): the prompt for this step
- tools (optional): list of tool names this step may call
- output_schema (optional): JSON schema (as a string) constraining the output
Unlike Agent / Summarizer, this node does NOT take a model property —
the workflow's configured model is used.
Properties
| Property | Type | Description | Default |
|---|---|---|---|
| instructions | str |
Instructions for the LLM step (acts as the user prompt). | `` |
| tools | list[str] |
Optional list of tool names the step is allowed to call. Empty = no tools. | [] |
| output_schema | str |
Optional JSON schema (as a string) constraining the step output. Empty = freeform text. | `` |
| input | any |
Upstream data forwarded to the step as context. Connect any node here. | null |
Outputs
| Output | Type | Description |
|---|---|---|
| output | str |
Related Nodes
Browse other nodes in the nodetool.agents namespace.