Tool

Node Overview

Node name: Tool (uses the name of the selected tool)
Display name: Tools (each node on the canvas displays its selected tool's own name)
Supported scenarios: Chatflow Agent (supported), Workflow Agent (supported)

Feature Description

The Tool node triggers a call to an external tool (both inside and outside Longbridge) and feeds the call result back into the flow for downstream nodes to use. It provides powerful tool integration capabilities, supporting calls to a variety of external tools and services.

How to add: click the node panel at the bottom left of the canvas, switch to the Tools tab (you can switch between the Infra / OpenAPI sources), find the target tool by category (News and Community Content, Real-time Quotes, Technical Analysis and Screening, etc.) or by search, and click it to add it to the canvas:

Node Display Content

1. Basic Information Display

• Icon: node-specific icon
• Name: the tool name (such as "Get News Details (Batch)")
• Note: explanatory text added by the user

2. Action Buttons

• More button: click to display the context menu content
• Run button: click to enter the test run interface

Node Edit Panel

1. Tool Description

• The top of the panel shows the tool name and a description of what it does (such as "Get News Details (Batch): fetch news details in batches by ID")

2. Input Variables

• Parameter definitions: displayed item by item according to the tool's input parameters—parameter name, type (such as array), whether required, and parameter description (such as "the list of IDs to fetch details for, maximum length 20")
• Value assignment: choose Variable to reference an upstream variable, or set a fixed value directly

3. Display Configuration and Citations

• Display configuration: a switch controlling how the tool result is displayed on the conversation side
• Citations: a switch controlling whether the answer carries citation sources for this tool's results

4. Output Variables

• text (string): the text content generated by the tool
• data (object): the structured content generated by the tool

5. Fault Tolerance Configuration

• Retry on failure: a switch; automatically retries when the call fails
• Exception handling: a dropdown to select the exception handling strategy (defaults to None)

6. Next Step Configuration

• Adding a next node is supported

Execution Logic

The execution logic of the Tool node is implemented by the backend to ensure tool calls are correct and secure.

Context Menu Actions

The context menu contains the following options:

1. Change node: change the node type
2. Copy: copy the node
3. Duplicate: duplicate the node content
4. Delete: delete the node
5. Help link: jump to the help documentation
6. Run node: test-run the current node

Test Run

• Support status: test run supported
• Change node: changing the node type is supported
• Copy features: copy, cut, and delete operations are supported

Tool Sources and Categories

• Sources: built-in Infra tools / custom OpenAPI tools (switch within the Tools tab of the node panel; for third-party MCP tools, see the Agent node and plan descriptions)
• Built-in Infra categories: News and Community Content, Real-time Quotes, Technical Analysis and Screening, Fundamentals and Financial Data, Macroeconomic Data, AI Analysis and Signals, Data Visualization Charts, Accounts and Positions, Order Management, Basic Tools and Search, User Memory and Personalization, Help Documentation
• The range of available tools varies by plan and region (built-in tools: partial on Starter, full on Pro and above)

Use Cases

1. Quotes and News

• Fetch real-time quotes and stock-related news / UGC content, then hand them to the LLM for summarization and interpretation (retrieve first, then answer—see Pattern 3 in section 2.3)

2. Technical Analysis

• Fetch candlestick and other data, then hand it to a Code node to compute indicators (do not hand large-volume computation to the model—see Tip 3 in section 2.5)

3. Account Scenarios

• Query account and position information, and order information (watch the compliance boundary when facing customers; this does not constitute investment advice)

4. Visualization

• Call chart tools to display visualized results on the conversation side

Configuration Steps

1. In the Tools tab of the node panel, add a tool by category or search
2. Configure the input variables (reference upstream variables or set fixed values; required items must not be missing)
3. Enable display configuration / citations, retry on failure, and exception handling as needed
4. Verify the result with a test run, then connect the next node

FAQ

Q: What if a tool execution fails?

A: Check whether the required input parameters are configured correctly and use a test run to view the error; enable retry on failure and configure an exception handling strategy.

Q: Why can't I find a certain tool?

A: The range of available tools varies by plan and region (e.g., only some built-in tools are available on Starter). See the Workspace and plan system documentation for details.

Q: How do I use tool output in downstream nodes?

A: text is suited to being referenced directly by LLM / Answer nodes; data, as structured content, is suited to processing and evaluation by nodes such as Code and IF Else.

Q: Are tool calls billed?

A: Only certain search and crawler tools are billed per call; the rest are free of charge. See the plan and usage documentation for details.

Considerations

1. Required parameters: unconfigured required input variables will cause the run to fail
2. Data authenticity: in your prompts, require the LLM to answer based only on the data returned by tools, with no fabrication
3. Compliance boundary: in customer-facing scenarios, watch output compliance (does not constitute investment advice)
4. Fault tolerance configuration: configure retry on failure and exception handling for important flows