# Call Application and Workflow Tool I/O Nodes

> Call sub-applications (runApp) in a workflow and configure pluginInput, pluginOutput, and toolParams for workflow tools.

Source: https://doc.fastgpt.io/en/guide/build/workflow/nodes/run_app

FastGPT lets you embed published applications or workflow tools inside other workflows for modular orchestration. This reference covers the sub-application node and the three dedicated system nodes used in workflow tools.

## 1. Call Application Node (`appModule` / `runApp`)

Run an existing Chat Agent, Chat Agent V2, or Workflow as a step on your current canvas, or attach it to a [Tool calling](/en/guide/build/workflow/nodes/tool) node for model-driven invocation.

### Inputs, Configuration, and Outputs

| Item                                   | Type             | Description                                                                                                                                               |
| -------------------------------------- | ---------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Target application**                 | Resource binding | Select a team application you have permission to use; the version selector in the node title bar offers "Keep the latest" or a specific published version |
| **User question** (`userChatInput`)    | `String`         | Input text passed to the sub-application from fixed text or an upstream variable                                                                          |
| **Chat history** (`history`)           | `ChatHistory`    | Context turns or upstream chat history passed into the sub-application                                                                                    |
| **File links** (`fileUrlList`)         | `Array<string>`  | File URLs passed to the sub-application                                                                                                                   |
| **Sub-app global variables**           | Dynamic fields   | Automatically exposes global variables declared by the target application                                                                                 |
| **AI Response Content** (`answerText`) | `String`         | Text response returned by the sub-application                                                                                                             |
| **New Context** (`history`)            | `ChatHistory`    | Updated chat history including the sub-application's turn                                                                                                 |

The node provides an **Error catch** switch to the right of the output section title. When enabled, sub-application run errors can be routed to a fallback branch; see [Debugging and Error Handling](/en/guide/build/workflow/debug-and-error-handling).

## 2. Workflow Tool System Nodes

When you create a workflow tool from the **Create** card in **Studio → Tools**, the canvas starts with **Plugin start** (`pluginInput`) and **Custom Plugin output** (`pluginOutput`) nodes in place of the regular start node; they contain the **Custom Inputs** and **Custom Outputs** sections. See [Create and Reuse Workflow Tools](/en/guide/build/tools/workflow-tools).

![Configure parameter variables in a workflow tool](https://cdn.jsdelivr.net/gh/labring/fastgpt-img@main/optimized/current/tool-variables.png.webp)

### 1. Plugin start (`pluginInput`)

* **Responsibility**: Declares the input parameter contract (JSON Schema) that callers (a parent workflow, Tool calling node, or Agent) supply when invoking the workflow tool.
* **Configuration**: Click **New** in the **Custom Inputs** section to add fields with a `key`, `label`, value type (`String`, `Number`, `Boolean`, `Object`, typed arrays, or `Select Dataset`), tool parameter description (guiding the model on how to populate the value), and required flag.
* **Outputs**: Each declared input field becomes a matching output handle on the node for downstream nodes to reference.

### 2. Custom Plugin output (`pluginOutput`)

* **Responsibility**: Collects results from internal workflow nodes and returns structured output fields to the parent caller or language model.
* **Configuration**: Click **New** in the **Custom Outputs** section to add output fields with keys, value types, and descriptions, and bind each field to an upstream node output (such as HTTP, Code run, or AI Chat).

### 3. Custom Tool Variable (`toolParams`)

* **Responsibility**: Declares runtime parameters to be generated dynamically by the language model for a sub-workflow connected to a [Tool calling](/en/guide/build/workflow/nodes/tool) node.
* **Availability**: Available when the canvas contains a Tool calling node or along a `selectedTools` tool connection. Declared fields are merged into the tool's Function Calling schema and exposed as node outputs at runtime.

## Example: Wrap and Call an Order Lookup Sub-Workflow

1. Create a workflow tool named `Order Status Lookup` and add a required `orderId` (`String`) field in **Custom Inputs** with the description `8-digit order number`.
2. Connect an HTTP node to query the order service and bind the returned status to a `statusText` field in **Custom Outputs**, then publish the tool.
3. Add the `Order Status Lookup` node to your main workflow, wire the extracted order ID into `orderId`, and reference `statusText` in downstream nodes.