← All guides

Replies you can use, not just read

Some answers are better as something you can use. Ask an agent for the open orders, the unread messages or this week's numbers, and it can put a table under its reply instead of a paragraph: sortable, searchable once it runs long, with a button on each row. It can also show cards, a short form, or a live view of the Proxy tab it's working in, so you can watch a change happen.

Buttons only ever run what the agent named when it made them, usually a capability you taught Proxy, filled in from that row. The reply can't point a button anywhere else later, and every press is recorded in the Audit Trail. Apps that come from other tools ask you first, in a box over the app, before they run anything at all.

In the desktop app a button can also take you somewhere: "View profile" on each row opens that profile in Proxy, signed in as the right account, and a button can run something and then land you on the page it finished on. If you've moved on by the time it finishes, a small "View result" message waits in the corner rather than pulling you away.

A table, cards or chart can also show a value the agent keeps in the workspace's saved values, rather than a copy of it. It updates on its own whenever the agent writes that value again, with no new message. Copy Shortcode, in the dots menu, gives you a short tag to paste into a workspace's prompt, and the same live view appears there with its buttons working: a small dashboard at the top of the workspace you check every morning.

Hide folds any of these down to a single line, so an old thread stays easy to scroll. They work in the desktop app, in a browser and on your phone; links into Proxy appear only where there's a Proxy to open.