Skip to main content
Version: v2.0

Permissions

An agent does not only answer questions. It sends messages, writes files, and changes records in the apps you connect to it. Permissions are how you decide in advance which of those actions it may take on its own, and which ones it has to check with you first. This page covers the answers an action can get, what happens when the agent checks with you, and why an unattended run needs the decisions made up front.

How permissions work​

Take an agent that reads records in a connected app and sends messages. You may want it to read on its own and check with you before anything goes out. Permissions make that split, and this example comes back throughout the page.

Every action an agent wants to take gets one of three answers before it runs: go ahead, ask you first, or refuse.

A tool can carry its own answer. Any tool that does not falls back to the agent's default, which is one of four:

  • Actions that only read run on their own, and anything that writes or sends asks you first. This is where a new agent starts.
  • Everything runs on its own, and you see what the agent did after the fact.
  • Everything asks you first.
  • Nothing runs.

The first default is the one the example agent needs: it reads the records on its own, and the send waits for you. You set the default once and it applies from then on without you being asked at run time. Control what an agent can do covers setting it and reviewing the actions your agent actually has.

When the agent checks with you​

An agent set to check first does the reading and the thinking on its own, and it stops at the moment it is about to act. It tells you what it wants to do and waits. Approve it and the run continues from where it paused. Turn it down and the action does not happen. For the example agent, that moment is the send: the reading is already done, and the message goes out only after you approve it.

An approval costs a second in a chat and stalls an unattended run​

What a stop for approval costs depends on whether you are there.

When you are chatting with the agent, it costs a second. You are already looking at the screen, you can read what the agent wants to do, and you can say no.

When the agent runs as an automation at six in the morning, nobody is there to answer. A request for approval has no one to approve it, and the run waits instead of finishing. For an unattended agent the decision has to be made up front: what it may do without asking, and what it must not do at all. Its instructions and the tools you give it carry the weight that you would have carried in a chat.

Next​