agenthub · function.tool
Function Tool
Use a tool you purchased from the marketplace. Connect it to an LLM or Agent Loop with a tool edge so the model can call it on its own.
Why the node earns its place
At run time this is identical to Tool Call — the same gateway, the same entitlement check, the same credential handling. The difference is authoring: the tool is a dropdown of what you have actually purchased, resolved when the palette loads.
Binding is explicit. A purchased tool that has not been dropped onto the canvas as one of these nodes is not available to the agent, which is what stops a marketplace purchase silently widening what every agent can do.
How it works at run time
- 1Bound with a tool edge: the contract is given to the model, which decides whether and when to call it, inside the LLM node's loop.
- 2Placed in the flow instead: it runs once, deterministically, like Tool Call.
What you wire
Inputs
- params *object
Outputs
- okboolean
- resultobject
- errorobject
- leg_costsarray
Configuration
| Key | Type | Default & options | What it does |
|---|---|---|---|
| tool_id * | string | choices from auth-layer:purchased_tools | A tool you've purchased from the marketplace. |
What usually goes wrong
Watch for this
Choosing the edge kind is the whole decision. Tool edge means the model may call it repeatedly and may not call it at all; a data edge means exactly one call, always, before the model runs.
Behaviour & provenance