SpaceXAI on Tuesday, August 11, introduced a new agentic AI product called Grok Bot that is designed to deploy a swarm of ‘always-on’ agents or ‘Bots’ to continuously carry out work across existing software and tools used by employees.Grok Bot has been jointly developed by SpaceXAI, the division of SpaceX formerly known as xAI, and AI coding startup Cursor, which was recently acquired by the space-tech giant for $60 billion. It is the Elon Musk-led firm’s latest bet on AI agents that move beyond answering prompts and toward completing long-running tasks.Currently available in early beta, Grok Bot allows users to spin up a team of autonomous agents or Bots, each with specific jobs. Upon giving them access to applications and websites, users can assign tasks for each Bot similar to how they would delegate work to a teammate, according to SpaceXAI.
Every Bot operates from within its own cloud-based computer environment and will get back to the user when it needs approval or has completed the task. The Bots remain functional even when the user’s laptop is closed.
The breakout moment for AI workflow agents can be traced back to OpenClaw’s release in November last year. Free for anyone to run on their computer, OpenClaw saw a surge in popularity among developers looking to build, deploy, and orchestrate agents or ‘claws’, which can further spin up their own sub-agents to execute specialised tasks.
Months later, the next frontier appears to be agents capable of long-horizon, more complex tasks.
Grok Bot is entering a crowded marketplace of first-party AI agents capable of automating enterprise workflows by interacting with applications on a user’s device. Several tech giants such as Google, Anthropic, and OpenAI have rolled out their own versions of agents that can do more than just generate text or code.Story continues below this ad
Still, Grok Bot could stand out as a more holistic management model for persistent AI agents, each with assigned tasks, enhanced memory, and learned user routines. The individual Bots are also capable of handing off work to one another.
In a post announcing the product, SpaceX said, “Bots are AI teammates that do real work for you. They sign in to your tools, use them just like you do, and come back with finished work.” “We will widen the Grok Bot beta after we fix basic issues with the early beta and release Grok 4.6 later this week,” SpaceX CEO Elon Musk added.
Grok Bot is available in beta from Tuesday for SuperGrok Heavy, Cursor Ultra, and Cursor Premium Teams subscribers. It is supported by macOS, Windows, Linux, and iOS devices, with Android support coming soon.
What is Grok Bot? What are its use cases?
SpaceXAI said that Grok Bot originated as an internal prototype, where teams created a swarm of Bots to automate sales outbound, marketing campaigns, office operations, bug fixes, and other work. It later decided to turn the internal tool into a full-fledged product for users.Story continues below this ad
Grok Bot is described as a team of ‘always-on agents’, where users can assign a role to each Bot and let them work simultaneously. Grok Bot has been developed keeping in mind large enterprise customers whose systems were likely never designed for AI-driven automation.
For instance, a sales Bot can be tasked with researching accounts, gathering prospective contacts, preparing email pitches and LinkedIn outreach in the user’s voice, and present combined results for human approval before hitting send. They can also add call-transcript notes to a CRM and draft follow-up messages.
On the other hand, an operations Bot can seat new hires and process invoices arriving through Gmail while an engineering Bot can reproduce a bug in the product interface, file a ticket, and then hand-off the repair work to a debugging Bot.
What are the alternatives to Grok Bot?
Anthropic first introduced computer-use for Claude back in 2024, allowing LLM-powered AI agents to navigate interfaces online. It continued expanding the tool with the launch of developer-focused Claude Code in 2025.Story continues below this ad
Earlier this year, Anthropic introduced its Claude Cowork agent focused on automating white-collar work. Its release sent shockwaves through global technology markets, including Indian IT stocks, with the episode being dubbed as ‘SaaSpocalypse’ in reference to software-as-a-service (SaaS) companies facing potential obsolescence due to AI disruption. Its broader Claude product also connects with workplace services and remote Model Context Protocol (MCP) servers.
OpenAI, on the other hand, updated its Codex harness with computer-use abilities. It also launched Workspace Agents that can connect to third-party applications and use them autonomously. More recently, the AI startup debuted a new ChatGPT Work environment for longer, multi-step tasks and finished deliverables.
How is Grok Bot different from other products?
SpaceXAI has not released Grok Bot’s performance on benchmarks for agentic AI tasks, making it hard to directly compare it with similar tools by rivals such as Anthropic and OpenAI. However, the following points differentiate Grok Bot from other offerings on the market:
– No API, MCP required: SpaceXAI has said that each Bot has its own computer and can continue working 24/7. It does not require every application or website to expose an API specifically for an agent. This means that automation is not limited to formally integrated services. It is designed to work through the same software interfaces a human employee would use.Story continues below this ad
– Proactive Bots over time: With Grok Bot, users can demonstrate how they complete a certain task and a Bot will follow along, save the process as a routine, and execute it later without requiring the user to reproduce every instruction. While a similar ‘Record the Skill’ feature is offered by Anthropic in Claude Cowork, SpaceXAI claimed that Grok Bot could go one step further with Bots, over time, identifying work before the user explicitly asks for it. Grok Bot also supports multiple agents operating together. They can be placed in the same thread, allowing the agents to pass messages, share context, and work amongst each other.
– Model agnostic: Interestingly, SpaceXAI’s announcement does not mention the underlying LLMs used to power Grok Bot. Additionally, there is no mechanism in place for users to select or switch to a particular model developed by SpaceXAI or a third-party developer. Instead, Grok Bot automatically routes tasks to models on the backend.
– Pricing: In comparison to Claude Cowork and ChatGPT Work, Grok Bot is priced competitively at $120 per seat per month for teams, and $200 per month for individuals. However, enterprises still face a difficult purchasing decision. The key question is whether persistent Bots can automate enough manual work and deliver productivity gains to justify their per-seat or per-user cost. Usage limits could also drive up costs when agents run continuously.
-Clean, simple interface: Grok Bot’s key differentiating feature is that it packages computer-use, multi-agent coordination, persistence, and workflow learning into a simplified tool that is capable of completing the work and putting it within the destination application.Story continues below this ad
The timing of Grok Bot’s debut is perhaps less than ideal as headlines over the past few weeks have been dominated by a string of cybersecurity incidents disclosed by OpenAI, Anthropic, Meta, and Moonshot AI, where under-testing or publicly available AI agents have broken out of containment, behaved in unexpected ways, and hacked into other platforms. The so-called ‘persistence’ of AI agents has also emerged as a troubling concern, as this recent bizarre incident demonstrates.