Originally published on LinkedIn. View the original post ↗
I recently added an MCP App workflow to one of my apps, specifically for the draft-then-approve flow I advocated for at All Things Open last week.
The agent prepares the work. Your product supplies the view of the data and the controls for the user to review, edit and approve the change, right inside the chat.
The clips below show Figma, Amplitude, Box and Slack bringing these richer experiences into the conversation.
There's some adoption here, but honestly, not nearly enough.
This keeps users engaged with your product while they work through their own assistant. You still control how the data is presented and how a draft becomes a committed change. Users never experience the friction of getting pushed to the browser, can witness and make their approvals in situ.
That hybrid of agent workflows, deterministic data review and controlled changes is something every agent surface needs to consider. I'm adding it to all my MCPs this week.

