Skip to content
AI agents & answers

Choose which tools an agent may use

Enable capabilities intentionally and keep operational automation separate from permission to use a tool.

Where to find itAI agents → agent → General / Experimental

Before you begin

Open AI agents and select the agent you want to edit. Check its name and workspace before saving: other agents have their own settings.

Example workspace · Demo data. Start with the correct agent and review the save controls. Select the image to enlarge it.
Agent settings showing identity, prompt and conversation options

Example workspace · Demo data. Start with the correct agent and review the save controls. Select the image to enlarge it.

Step by step

  1. 1

    Review the task

    Decide which capabilities the assistant actually needs, such as site lookup, commerce or staff handoff.

  2. 2

    Enable the necessary tools

    Use the agent’s tool controls. If a capability needs a store or channel connection, complete that connection before testing.

  3. 3

    Check restrictions

    A workflow can narrow available tools for a case, but cannot grant tools that the agent or connection is not permitted to use.

  4. 4

    Test an allowed and denied case

    Verify the tool succeeds with valid context and that the assistant responds honestly when access, data or identity evidence is missing.

Good to know

Experimental commerce controls can affect real visitors. Use a safe test case and verify customer identity requirements before order-related actions.

Check the result

Save, start a fresh test conversation and compare the answer with your approved business facts. Settings do not rewrite messages already sent.

Questions & troubleshooting

The model mentions a capability but cannot execute it

Check saved tool permissions, the connection, agent assignment and any workflow restriction. A prompt alone does not enable tools.

Was this article helpful?

Rate this guide on this device. Ratings are not sent to Corthex.