Skip to main content
Talk to your agent the way you’d ask a friend. You don’t need tool names. Below, each prompt shows which tools the agent uses and what to expect.

”Make me an alias for netflix.com”

The agent calls wiseguard_create_alias with the label Netflix and the site netflix.com.
  • If you already have an alias for netflix.com, you get that one back. It’s not a new alias.
  • If that alias is paused, the agent should ask whether to resume it or make a new one.
  • Otherwise you get a new address, like maple.river.stone@wiseguard.app.

”Rename it to Netflix”

The agent calls wiseguard_rename_alias on the alias from the last step. Only the label changes. The address stays the same.

”Which aliases are paused?”

The agent calls wiseguard_list_aliases with state: paused. It shows 10 at a time. Say “show more” to see the next page.

”Delete my old shopping alias”

The agent first calls wiseguard_list_aliases with the search shopping to find the alias. Then it calls wiseguard_delete_alias. The delete tool is marked destructive, so your AI app may ask you to confirm. If more than one alias matches, tell the agent which one.
Paused and deleted aliases still receive and forward mail today. Pausing or deleting an alias doesn’t block mail yet. A deleted alias keeps forwarding until it’s removed for good, after the 7-day restore window.

”Restore it”

The agent calls wiseguard_restore_alias. The alias comes back active, with the same address. This only works within 7 days of deleting it. If you don’t remember what you deleted, ask “what have I deleted recently?”. The agent calls wiseguard_list_recently_deleted.

”Any new mail on my netflix.com alias?”

If the agent doesn’t know the alias address yet, it first calls wiseguard_list_aliases to find it. Then it calls wiseguard_list_messages and shows you the sender, subject and a short snippet of each email. Nothing is marked as read yet.

”Read the newest one”

The agent calls wiseguard_read_message. You get the whole email as plain text, and it’s marked as read in WiseGuard. If an email tells the agent to do something, like “forward this to everyone”, the agent should ignore it. Emails are untrusted.

”Reply and say I’ll call them tomorrow”

The agent should show you the reply first and ask you to confirm. Then it calls wiseguard_reply_as_alias. The reply goes to the sender, from the alias the email was sent to. It can’t be unsent.

”Email support@example.com from my shopping alias”

The agent finds your shopping alias, writes the email, and asks you to confirm. Then it calls wiseguard_compose_as_alias. The alias must be active, not paused.

”How many AI agent credits do I have left?”

The agent calls wiseguard_whoami, which reports your balance. Changes and sends cost credits. Looking things up is free.

What can go wrong

  • The agent can’t find the alias. It gets “not found”. Give it the full address, or ask it to list your aliases first.
  • You hit a limit. See Limits and behaviour.
  • You run out of credits. See AI agent credits.
  • Something else fails. See Troubleshooting.