xAI has launched Grok Bot, a system of autonomous AI agents designed to work across applications and tools without continuous user oversight. The company announced the beta release on Tuesday, positioning the product as a team of AI teammates that can handle multi-step tasks end to end.
Each Grok Bot operates from its own cloud-based computer and can sign into the tools a user already relies on, including platforms that lack clean APIs or Model Context Protocol support. The agents work across apps, inboxes, and websites, returning only when a task requires human approval.
- xAI launched Grok Bot in beta on Tuesday
- Each agent has its own cloud computer and can sign into existing tools
- Agents work across apps and websites, including those without APIs
- Available for SuperGrok Heavy, Cursor Ultra, and Cursor Teams Premium subscribers
- Enterprise users can join a waitlist for future access
- Bots can learn workflows by watching users and save them as reusable routines
- Multiple Bots can coordinate in parallel through group chats
Grok Bot began as an internal prototype at SpaceXAI, where teams created agents for sales outbound, marketing campaigns, office operations, and engineering tasks. The company said adoption spread quickly across the organization before the decision was made to open it to external users.
Inside SpaceXAI, the company described several concrete use cases. A sales Bot updates customer relationship management systems with call transcript notes and drafts follow-up messages. An operations Bot processes invoices received in Gmail and handles seating arrangements for new hires. An engineering Bot reproduces bugs in the product interface, files tickets, and hands fixes off to a debugging Bot.
Users interact with Grok Bot through messaging, either from a phone or desktop, similar to texting a colleague. Conversations persist across devices, and the agents remember context from previous interactions. Over time, a Bot learns how a particular user prefers work to be done by observing them complete a task, then saves that sequence as a reusable routine that runs automatically on subsequent requests.
The learning mechanism works by having a Bot follow along as the user completes a job. It watches the steps, remembers preferences, and stores the workflow. The user can correct the Bot’s understanding, and the agent incorporates those corrections before running the routine independently the next time the same type of task arises.
The system supports multiple Bots running in parallel, managed by what xAI describes as a chief of staff Bot that oversees specialist agents focused on specific functions such as inbox management, expenses, recruiting, bug fixes, or general operations. Bots can message each other independently and share context within threads, which the company said allows them to stay aligned on overlapping projects without requiring the user to manually transfer information between conversations.
Bots can also be placed in group chats where they coordinate on their own, passing work between agents, assigning ownership, and pulling the user in only for judgment calls. xAI said this structure is designed to prevent the user from becoming a bottleneck when multiple tasks need to move forward simultaneously.
Over repeated use, Bots become more proactive. xAI said they eventually begin picking up work before being explicitly asked and learn to recognize when something requires human attention versus when they can proceed independently.
Grok Bot is available in beta for subscribers of SuperGrok Heavy, Cursor Ultra, and Cursor Teams Premium on desktop and iOS. Enterprise users can join a waitlist for future access. xAI did not specify when the product will reach general availability or what pricing will apply outside the current beta tiers.
The Cursor subscription tiers listed as eligible for Grok Bot access reflect the deepening relationship between the two companies. xAI signed a deal with Cursor earlier this year that gives the code editor access to xAI’s Colossus supercomputer.
The launch follows xAI’s earlier entry into the agentic AI space with Grok Build, a coding agent released in May that works from the command line interface. Grok Bot extends the company’s agent offerings beyond software development into general business operations, placing it in competition with agentic tools from Anthropic, OpenAI, and other AI labs that have been racing to deploy autonomous systems capable of acting across software environments.