Typing a good prompt is harder than it looks. If the task is visual, like a misaligned button or a confusing user flow, you end up pasting screenshots, describing arrows in words, and answering follow-up questions from an agent that still got it wrong. Atlassian thinks the fix is something Loom users already do every day: just record it.
The company announced a bundle of Loom updates built around that idea. The headline is Record for agent, joined by a rebuilt Loom for Mac, Loom tools inside Atlassian MCP, and Rovo in Loom.
What Record for agent actually does
You record your screen and talk through the task. Loom captures the transcript, screenshots of the elements you clicked, and the links you visited. It then packages all of that into a prompt an agent can act on. You can send it straight from Loom or open it as a Jira work item and assign it to an agent.
The new Loom for Mac is a rebuilt desktop app that records in 4K by default. It adds two tools aimed at this workflow: a Selection tool that pins a specific UI element, and a Magic pen for freehand annotations. Atlassian pitches use cases like UI fixes, slide decks, summaries, prototypes, and feature specs.
The MCP side matters more
Record for agent is the flashy part. The MCP integration is the part that makes it useful beyond Atlassian’s own tools. Loom tools are built into Atlassian MCP, so agents like Claude, Cursor, VS Code, Codex and Rovo can pull a video’s transcript, metadata and action items, add timestamped comments, and organize recordings. Permissions follow the team’s existing Loom setup, and there’s no separate server to configure.
If you’re on Atlassian MCP v2 and your Loom workspace is linked to your Atlassian site, it’s available automatically. v1 users need to update or reinstall the connector.
Rovo in Loom is also rolling out. Rovo Search finds a video or moment across a library, and Rovo Chat answers questions, drafts follow-ups, or turns a walkthrough into written steps, with answers tied to exact timestamps.
Pricing and availability
Record for agent is in open beta for Loom Business + AI and Enterprise plans. It works in the new Loom for Mac and the Chrome extension. Smaller plans are left out for now.
Does it matter?
Partly, yes. The problem is real. Atlassian cites its own State of Teams 2026 report, where 87% of knowledge workers say they lack the time or capacity to coordinate. Visual context is a genuine gap in most agent workflows, and coding tools like Cursor and Claude Code mostly still rely on text and pasted images.
But there are caveats. The pitch leans on Atlassian’s own survey data and a Fortune 500 CTO quote, not independent evidence that video prompts produce better agent output. A transcript plus screenshots is also something other tools could assemble, and Loom’s real edge is distribution: it’s already in many teams’ habits, and it ties into Jira and the Teamwork Graph.
That’s the strategic play. Atlassian wants Loom to be the context layer agents read from, not just a video tool humans watch. If you already pay for Loom and Jira, try it on a UI bug. If you don’t, this is a nice feature, not a reason to switch.
