Skip to main content

AI chat assistant

The chat assistant (sidebar → Chat) is a general-purpose AI surface that spans all your workspaces. It can answer questions about your accounts and analytics, draft captions, and — unlike a plain chatbot — act: it can schedule posts, attaching media it picks from your library.

Chat runs on your brand profile and per-account voice like every other AI surface, so replies already sound like you.

What it can do​

The assistant can call these tools on your behalf:

  • list channels — see every connected account.
  • schedule post — create a scheduled post on a channel you name, with a caption (extracted from the conversation if you don't repeat it), optional first comment, and media by library id. Times must be in the future.
  • get recent posts — look up what was recently published.
  • get analytics — pull post counts by status and recent activity.
  • create clipping — turn a public YouTube video into captioned vertical clips and optional drafts for selected channels. See video clipping.
  • get clipping / list clippings — check progress and retrieve finished clips.

Ask naturally: "schedule a post with the sunset photo on Instagram for Friday 9am", "how did my accounts do this week?", "write 5 hook variations for a LinkedIn post about our launch".

Each message draws from your plan's monthly AI-action allowance. When the allowance is exhausted, chat tells you the limit and when it resets (first day of next month, UTC).

Threads​

  • Conversations live in the left sidebar; New Chat starts a draft that is only persisted once you actually send a message — an abandoned draft never clutters the list.
  • Threads are auto-named from your first message. The list pages 20 at a time.
  • Select mode lets you select all / several threads and delete them in bulk (with confirmation). Deleting the open thread returns you to the empty state.
  • Deep links work: /chat?threadId=<id> opens a specific conversation.

Images​

Two ways to work with images:

  • Attach — upload a file (images only, ≤10 MB; HEIC is converted automatically) or pick from your media library to ground the conversation. Attachments are described by vision once and cached, so follow-up questions about the same image don't re-burn allowance.
  • Generate — switch to image-generation mode from the + menu, describe what you want, and the result lands in the conversation and can be saved straight to your library. Generations draw from your monthly image-generation allowance, separate from AI actions.

While the assistant is streaming a reply you can hit Stop to abort it.

Under the hood​

For the curious: each send assembles your brand context, per-channel voices, the thread's media inventory, and the last 40 messages; the assistant may run up to four tool rounds per turn. Streaming responses arrive token-by-token over SSE, and messages render as Markdown.