Atlassian rebuilds Loom for Mac and lets recordings brief AI agents

You have probably opened a screen recorder, glanced at your desktop and realised that a sea of loose files and a camera bubble are about to star in your product walkthrough. The bubble sits over the exact button you need to click, and you spend the first five minutes rehearsing bullet points in your head. Atlassian wants to remove that friction for people explaining work to teammates and, increasingly, to AI agents.

What Atlassian rebuilt

Atlassian retired Loom's legacy desktop foundation and rebuilt the Mac app natively for macOS 15 and later using Apple's own technologies. The app opens faster, responds more quickly and uses less memory and processing power while it records. It is also about 80 per cent smaller, which cuts the download and the disk footprint. The rebuild taps more advanced macOS screen-capture features and defaults to higher resolution, with support for 4K video, richer colour and clearer text. The team also reworked camera and microphone handling so switching devices is easier and backgrounds and filters run with less lag.

The most important layer is the least visible. Loom now has a screen-awareness foundation that tracks the screen, windows, mouse movement and recording area in real time, and that layer powers most of the new features. Loom uses it to separate the content you mean to share from the surrounding Mac chrome, automatically hiding desktop icons and the top menu bar from the final recording, so bookmarks, open tabs and background notifications stay private without a manual clean-up. It also lets users drag the camera bubble around the edge of the screen and snap it into place without covering critical on-screen elements. New squircle, portrait and landscape bubble shapes join the classic circle, alongside one-click filters and backgrounds that include gradients, brand assets and custom colours. Drag-to-select lets presenters capture a precise area, and they can switch between full screen, single windows, custom areas and camera-only mode.

Speaker notes and the teleprompter

Loom has refreshed its speaker notes and built-in teleprompter so presenters can stay on message without breaking eye contact. Users can position the scrolling teleprompter just below the Mac camera, and voice-to-text turns rough ideas spoken aloud into structured speaker notes. AI Speaker Notes, powered by Atlassian's Rovo, turns rough thoughts or bullet points into clear, structured talking points in seconds.

Recording for agents

The feature with the biggest implications is Record for agents. Atlassian's argument is that prompting AI should work the way you brief a coworker: you pull up your screen, point at the issue and narrate your thinking. With Record for agents, a user records the screen, clicks through the relevant links or design files and talks through the task. Loom captures the voice, clicks, cursor movement and on-screen context, then turns the recording into a step-by-step action plan that agents such as Rovo, Claude or coding assistants can execute. Magic Pen and Key Moments let the presenter circle specific interface elements and flag important context during the recording, such as new design requests for a landing page, which feeds precise technical metadata and screenshots into the instructions. When the work is ready to move, Loom can convert recorded action items into structured Jira issues with a single click.

What this means for L&D and distributed teams

Learning and development teams have long traded off live training against self-serve documentation. Live sessions offer real-time questions but struggle across time zones, and written guides scale but often lose the nuance of a software workflow that a two-minute screen recording captures easily. Automatic decluttering, smarter bubble placement and a teleprompter lower the production overhead, so a subject-matter expert can record a clean walkthrough without treating it like a film shoot.

Record for agents opens a different door for innovation and transformation leaders. Teams experimenting with agents need a way to hand off context that is richer than a chat message and faster than a written spec. A narrated screen recording that becomes an action plan an agent can execute, and then a set of Jira issues, turns the handoff into something people already know how to do. The capability to build is the briefing itself. Teach people to narrate a task clearly, mark the critical context as they go and check the action plan before an agent runs it, because a vague recording will produce a vague plan no matter how sharp the 4K capture is.

Sources:

• Atlassian, 7 October 2026