Skip to main content

Tools

A custom Tool packages a focused AI task so a Workflow can call it when needed. Its description tells the calling Agent when to use it; its prompt tells the Tool how to complete the task with its available Abilities.

Create a Tool​

  1. Open Settings → Toolkits → Custom Toolkits. On the Toolkit that should own it, choose Add Tool.
  2. Enter a Name, Display Name, and Description. Use a function-style Name such as summarize_request; it cannot be changed after creation. Describe the request this Tool accepts and the result it returns.
  3. In Chat, write the Prompt. State the task, what information to use, and what to return. Select a Chat Model if the task needs a specific one.
  4. Configure Inputs and Outputs. Enable Use Custom Input Fields or Use Custom Output Fields when the caller needs named, typed fields instead of plain text; define those fields in the editor.
  5. In Abilities, add the capabilities required to complete the task. Review the Abilities inherited from the Toolkit as well.
  6. Save the Tool.

A prompt alone does not connect a service or grant access to its data. Configure the relevant App Connection and select the appropriate Ability when the task requires one.

Check the result in a Chat​

Attach the Toolkit to a Workflow or Task, then test a request that should call the Tool. Check that it receives the intended inputs and returns a result the Workflow can use. Include a missing-input case so the Tool's instructions explain how to ask for what it needs.

Use the Toolkit's Tool list to edit or delete the Tool. Before deleting one, update Workflows that depend on its result.

For API definitions and structured Tool contracts, see Developer Toolkits. For reviewing the main assistant's behavior, use a Supervisor.