Permissions
Switch individual agent capabilities on or off, independent of what resources are attached.
The Permissions tab lets you switch individual agent capabilities on or off with simple toggles. These control what an agent is allowed to do, separately from what resources are attached to it elsewhere in the editor — an agent can have a knowledge base attached and still be blocked from certain actions if the matching toggle is off. There are around 16 of these fine-grained toggles in total; the table below covers the main ones.
| Capability | What it unlocks |
|---|---|
| Create or modify skills | Build new skills or edit existing ones, not just use them |
| Modify knowledge bases | Add, edit, or remove entries in a knowledge base, not just search it |
| Use memory | Save and recall notes about you and the workspace across conversations |
| Plan and execute multi-step tasks | Break a request into steps and carry them out without asking at each step |
| Configure integrations | Set up or change connected accounts itself |
| Read user info | Look up details about you or your workspace profile |
| Use data storage | Store and retrieve structured data it generates |
| Use web search | Search the web for current information |
| Generate images | Create images in a conversation |
| Work with PDF templates | Select and use the PDF templates attached to it |
| Generate documents | Produce documents, such as filled templates or exports, as output |
| Modify workflows | Create or edit workflows, not just run the ones attached to it |
Adjust these to match what you want a particular agent to be able to do — for example, turning off Use web search and Generate images for an agent that should only answer from its knowledge base.
Some capabilities are granted automatically when you attach a matching resource — see Assigning Resources for the knowledge base example.