SpaceXAI has introduced Grok Bot in early beta, an AI agent built to work directly inside apps and websites and handle multi-step tasks with minimal human supervision.
Unlike traditional AI assistants that mainly provide answers or instructions, Grok Bot can operate software on behalf of users. Each bot runs on a persistent cloud-based virtual machine equipped with its own browser, filesystem, and terminal. It can also use connected tools and computer-use capabilities, allowing it to interact with applications even when they do not offer an API.
Users can create separate bots for tasks such as managing emails, researching potential customers, tracking expenses, or reproducing software bugs. Multiple bots can operate simultaneously and share context, enabling one agent to hand off part of a task to another.
Grok Bot can also learn workflows from user demonstrations. Once a user shows the bot how to complete a process, it can save that workflow as a routine and repeat it later, either manually or on a schedule. Bots can retain their files, browser sessions, preferences, and previous context between tasks.
SpaceXAI says the goal is for Bots to complete tasks from start to finish and only involve the user when approval or another decision is required. For example, a bot could research potential customers, evaluate accounts, and prepare outreach messages, while leaving the final communication for the user to approve.
The early beta is currently available to SuperGrok Heavy, Cursor Ultra, and Cursor Premium Teams subscribers on desktop and iOS. Enterprise users can join a waitlist for access.
The Cursor integration follows SpaceXAI’s reported agreement in June to acquire the AI coding company in an all-stock deal valued at around $60 billion. Before the acquisition, Cursor and SpaceXAI worked together to train Grok 4.5, using trillions of tokens from developer-agent interactions. The model was designed to support long-running tasks extending beyond traditional software development.
Also Read: SpaceXAI Launches Grok 4.5, Narrowing the Gap with GPT 5.5 and Opus 4.8
