Skip to main content
A thread holds the conversation, the tool runs, and the outputs for one piece of work. The composer is how you send new work into it. This page covers what the composer can do and what you see while a thread is running.

​
Write into the composer

The composer is a plain text box, with two shortcuts that pull in more than text:
  • Type / to insert a skill. A list of the skills available to you appears, and picking one gives the thread reusable instructions for how to approach the task. See Skills.
  • Type @ to insert a connector or a file. A connector gives the thread access to a system it needs; a file gives it something to read. See Tools and connectors and Files and artifacts.
You can also paste or drag a file straight into the composer to attach it, without typing @ first.

​
Watch the work happen

1

The reply streams in

Once you send a message, the reply streams in as it is produced. You don’t wait for the whole answer before you can start reading it.
2

The activity ledger fills in

Alongside the reply, the thread keeps an activity ledger of what is happening: which tools were called, what they returned, and any approvals that were asked for. This ledger stays in the thread after the work finishes, so you can come back and see exactly what ran.
3

Stop if you need to

If the thread is heading somewhere you don’t want, press Stop. This ends the current run. You can then send a new message to redirect the work or pick up where it left off.
If a tool call needs your sign-off, the thread pauses on its own and shows an approval card instead of continuing to stream. See Approvals for what to do with it.

​
Where to go next

Skills

Reusable instructions you bring into a thread with /.

Tools and connectors

Give a thread access to the systems it needs.

Files and artifacts

Upload context and collect the outputs a thread produces.