Mentions
Type@ in the composer to open an autocomplete menu and pull in sources or skills inline:
- Sources — mentioning a source (for example
@githubor@linear) activates it for the conversation and grants the agent the tools it needs to work with that service. - Skills — mentioning a skill (for example
@daily-standup) invokes a packagedSKILL.mdinstruction set so the agent runs a reusable procedure on demand.
File attachments
Share files with the agent by dragging them into the composer, pasting from the clipboard, or using the file picker. Synti handles a broad range of formats, and the agent can read, review, extract from, analyze, and transform them:How responses stream
When you send a message, the agent responds by streaming its work in real time — reasoning, tool calls, and output appear as they happen. You can watch each step, and stop the response at any point if it’s heading the wrong way.Steering vs. queuing
You don’t have to wait for the agent to finish before you type again. Two modes govern what happens to a follow-up sent while the agent is still working:- Steer — your new message folds into the current turn before it completes, letting you redirect the task mid-stream.
- Queue — the current turn finishes on its own, and your follow-up starts a fresh turn.
Steering is useful for fast course corrections; queuing is better when you want each turn to complete cleanly before the next instruction lands.
Managing conversations from the thread
While you interact, you can also keep the conversation organized: flag important threads, archive inactive ones, and act on several at once with batch operations like archive, delete, or label. New agent output marks a conversation as unread so you can see at a glance which threads need attention. See Conversations for the full inbox model.Keyboard shortcuts
Common actions have keyboard shortcuts, and the bindings are customizable:On Windows and Linux, use
Ctrl in place of Cmd.