Skip to main content
A support agent you bring works customer threads the way a person on your team does. It picks up a thread, reads it, and replies, labels it, leaves a note, or hands it to someone. Because it acts as a machine user, your reporting attributes its work to it: volumes, resolution times, and response times all count it the same way they count a person. You decide how much it does on its own. An agent can draft replies for a person to review and send, which is a good place to start, or it can answer autonomously and hand off when it cannot help.

​
What your team sees

Threads your agent is working stay out of the First Response, Next Response, and Investigating queues, so your team only sees what needs a person. Everything the agent touched is visible under Plain → AI → Agent Activity.

​
Routing work to it

Assign threads to your agent’s machine user and it works only those. A workflow can do the assigning on a trigger and conditions you choose, such as channel, labels, customer tier, or support hours, so you change what your agent handles without touching code. Anyone on your team can hand a thread over by reassigning it, and the agent can hand back the same way.

​
Where to go next

The technical detail lives in the developer docs:

Build a support agent

Identity, webhooks, routing, reading a thread, and the actions your agent takes.

Building agents on Plain

How an agent gets an identity, receives events, and decides when to act.