Tools & integrations

By the end of this page you will be able to call a tool from inside your Task Agent workflow.

Prerequisites

A Task Agent created (Create a Task Agent).
Owner, Admin, or Developer role.

Where tools come from

Tools are created and managed in their own section — Agentic Workspace → Tools — because they’re shared across agent types. This page covers how you use one inside a Task Agent. To build or manage tools, see:

Tools overview — scopes (System, Project, MCP), categories, lifecycle
Create a tool — the create wizard (HTTP · Python)
Configuration & secrets — reference API keys safely with {{config.xxx}}

The Tools Call node

Inside a Task Agent, you invoke a specific tool with the Tools Call node. Unlike an AI Agent — where the model chooses tools — a Tools Call node is deterministic: you pick the tool and wire its inputs, and it runs exactly where you placed it in the graph.

Add the node

On the canvas, add a Tools Call node from the Actions category of the node palette (keyboard shortcut T). Its description reads “Execute a system or project tool.” A new node shows “Select a tool.”

A Tools Call node on the canvas showing the Select a tool prompt

Steps

Select a tool

Click the node to open its panel. The left sidebar lists tools, split into two scope tabs:

System — built-in qRaptor tools.
Project — your published custom tools.

Search with Search tools… and narrow by category (All Categories, Function, Data Pull, Action). Click a tool to select it — a check mark appears. If you need a new tool, Create Project Tool opens the create wizard in a new tab.

The Tools Call node tool selector sidebar with System and Project tabs and search

ℹ️

Task Agent nodes expose a curated set of System tools (web, database, RAG, DMS, media, documents, and utilities). Some tools are AI-Agent-only and won’t appear here.

Configure the inputs

On the Configure tab, the Parameters section lists the selected tool’s inputs. Required parameters are marked. Fill each value directly, or reference workflow data with {{variable}} — type into the field or use the variable picker.

The Configure tab showing the tool parameters wired to workflow variables

Set execution and output options

Execution Settings:

Timeout (sec) — 1–300 seconds (default 30).
Retry on Failure — toggle, with a Max retries count.

Output Settings:

Output Variable — the name downstream nodes reference (default tool_result).
Include execution metadata — attach run metadata to the output.

Test the node

Switch to the Test tab, provide Test Input for each parameter, and click Run Test. The result shows Test Passed or Test Failed, the duration, any validation errors, and the Output.

The Test tab showing a passed tool run with output

Click Save to keep the node. The Save button is disabled until a tool is selected.

Using the output

Downstream nodes read the result as output.{'{'}nodeId{'}'}.{'{'}outputVariable{'}'} — for example output.toolsCall.tool_result. When metadata is included, the output carries both the result and metadata.

💡

Chain multiple Tools Call nodes to orchestrate complex workflows — for example, fetch data with an HTTP tool, transform it with a JSON tool, then pass it to a Quick Prompt node for summarization.

Google OAuth tools

Some tools — email reader, attachment download/list, and the Google Meet tools — need Google OAuth. If you select one without connecting an account, the node shows: “Google OAuth authentication is required. Connect your account in the Config tab.”

Common issues

“Select a tool to continue” — no tool is chosen; pick one from the sidebar before saving.
A tool you expected isn’t listed — it may be AI-Agent-only, or (for Project tools) not yet Published. See Publish & lifecycle.
Empty or wrong output downstream — check the Output Variable name and that you referenced output.{'{'}nodeId{'}'}.{'{'}outputVariable{'}'} correctly.
Validation errors on test — a required parameter is missing or the wrong type.