xAI Introduces New Grok Bot Design for Persistent AI Agents

xAI redesigned Grok Bot around persistent AI agents, giving each Bot memory, its own computer, and scheduled Routines that work without user prompts.

maisiekooc
Maisie Morrison

AgentLocker Editor

AI News
xAI Introduces New Grok Bot Design for Persistent AI Agents

Grok Bot has a new design built around agents that keep working after a chat session ends. The team behind it rethought how people track, monitor, and manage AI agents that carry tasks over days or weeks.

The designers started by asking which concepts users actually need to work with an agent. They narrowed a long list of technical terms down to five main ideas.

Those five ideas are Bots, Chats, Prompts, Tools, and Artifacts. Bots are the persistent agents themselves, each with memory, an identity, and its own tools.

From Chat History to a Bot Roster

Most AI products organize themselves around chat history. Older conversations get pushed down a list and users rarely return to them.

xAI decided this setup did not fit an agent meant to remember past work and carry responsibility over time. So the main screen in Grok Bot now shows a roster of Bots, not a list of chats.

Each Bot has a name, an avatar, and a title. It remembers previous conversations with the user. Coming back the next day means returning to the same Bot, not starting a new thread.

To make the roster easy to scan, designers built a visual system for Bot avatars. The avatars use simple shapes and expressive eyes so people can recognize a Bot at a glance.

The avatars also show what a Bot is doing. Motion changes depending on whether a Bot is idle, working, waiting, or finished with a task.

Bots Get Their Own Computer

Each Bot in the new design has its own computer. It can browse the web, open files, and run software on that computer.

xAI tested four ways to show this computer to users, including a floating window, a side by side view, a modal, and a full screen mode. Each option changed how much users felt pulled into supervising the Bot.

The final design settled on three access levels. A status indicator shows when the computer is active, a preview panel lets users watch without leaving the chat, and a full takeover mode lets users step in and control the machine directly.

Designers also added wallpapers that shift from light to dark through the day. This detail is meant to give each Bot's workspace its own sense of time, separate from the user's own desktop.

Beyond the Bot's screen, the interface changed how it displays information generally. Instead of describing a five-day weather forecast in text, for example, Grok Bot can show it as a structured card.

The same approach applies to actions like changing a setting or messaging another Bot. Those events now appear directly in the chat transcript as objects users can open for more detail.

xAI also built a system for coordinating multiple Bots. Some users created a "Chief of Staff" Bot that directs several other specialized Bots, cutting down on manual routing of tasks.

Each Bot keeps its own memory tied to its role, while shared tools like web browsing or email remain available across all Bots on an account. Group chats let several Bots collaborate on a project while keeping separate memories.

A new feature called Routines lets a Bot start work on a schedule or in response to an event, rather than waiting for a prompt. This means tasks like a daily briefing can run automatically.

xAI capped the system at roughly 50 Bots per account and six Bots per group chat. The company said each design choice came down to whether it helped users delegate work or added another thing to manage.

From our research desk
AI Jobs Automation Index
Which jobs are AI tools targeting most? We mapped 3,400+ AI tools to real job functions — with BLS employment & salary data.
Explore the index
maisiekooc

Written by

Maisie is a news writer at Agent Locker, covering the latest developments in artificial intelligence, emerging technology and the companies shaping the future.

Discover AI Agents