Agents
Prepare access for your Agent
Create a named key, choose its permission and understand how to stop its access.
The Python examples on the Agents page use an account key. Before connecting an Agent, prepare that access in Tokens and Connections. A useful key has a recognizable name, the permission your task needs and a clear owner.
Give the key a recognizable name
Open Tokens and Connections in Settings and choose the control for creating a token. Use a name that identifies its purpose, such as “Agent test”, instead of a name you will not recognize in the list later.
The illustrated key on the Agents page is a placeholder, not a credential you can use. Your created token is the access associated with your account. Keep it separate from any example or screenshot you plan to share.
Choose the permission for the task
Read the permission descriptions before creating the token. Execute is intended for running workflows and agents. Read allows viewing data; Write allows creating or changing it; All grants the full set of permissions.
For an Agent that only needs to execute a task, start from the Execute description. Do not choose All just because it is the broadest option. If the task also needs a different operation, review the permissions required by that operation.
Review access after your test
Find the token in the list by its name. Its permission and usage information help you identify the key you are using. A token marked as unused does not establish that your Agent call succeeded; check the call result separately.
The page distinguishes unpairing a device from deleting the token. Unpairing does not disable the key. To stop access through an old test key, delete that token using the confirmation shown by the page. Do not delete a key another active client still needs.
Cyber Academy
What to remember
- Name each key so you can recognize its purpose.
- Choose the permission that matches the operation.
- Unpairing a device and deleting a token do not have the same effect.