SpaceXAI has launched Grok Bot, a beta service that lets users assign ongoing work to AI agents that can operate apps and websites through their own computer environments. The release comes alongside Grok 4.6, a new model that SpaceXAI positions for coding, knowledge work and longer agent-led tasks.
Grok Bot is initially available on macOS, iOS, Windows and Linux. Android support is planned. Access is limited at launch to SuperGrok Heavy, Cursor Ultra and Cursor Premium Teams subscribers. According to SpaceXAI’s product information cited by VentureBeat, Cursor Ultra costs $200 per month for individuals, while the team plan costs $120 per seat each month. Existing SuperGrok Heavy subscribers also receive access.
The central idea is to move from one-off chatbot prompts to persistent assistants with defined responsibilities. Each Bot receives its own computer environment, signs in to services that the user already uses, and can keep working when the user’s laptop is closed. SpaceXAI says the agents return when a task is complete or when human approval is needed.
Users can assign Bots jobs such as sales research, recruitment, expense processing, customer support, marketing, CRM updates or software bug reproduction. The system is designed to operate software interfaces directly, including tools that do not offer a suitable API or Model Context Protocol connection.
From instructions to repeatable routines
SpaceXAI says users can demonstrate a process while a Bot observes it. The Bot can then save that process as a routine and repeat it later. The company also says Bots can learn from corrections, retain conversation context and adapt to preferences such as a user’s writing style.
Grok Bot supports multi-agent workflows. Several specialized Bots can share a conversation, hand tasks to one another and coordinate work. A Chief of Staff Bot, for example, could direct research, communications and travel Bots. This design could reduce the need for users to manually move information between different AI tools.
However, the product’s usefulness will depend on execution quality and controls. An AI assistant that drafts an inaccurate answer is one problem. An agent with access to customer records, invoices, inboxes or production software can cause more serious operational issues. SpaceXAI has not published independent benchmarks for Grok Bot itself, and its public materials provide limited detail on permissions, model selection and audit controls.
Grok 4.6 targets agent performance and lower inference costs
One day after the Bot launch, SpaceXAI released Grok 4.6. Artificial Analysis gives the model a score of 61 on its Intelligence Index, placing it level with OpenAI’s GPT-5.6 Sol Max in that ranking and behind Anthropic’s Claude Opus 5 and Claude Fable 5.
Artificial Analysis reports strong results for Grok 4.6 in multi-step agent tasks, including knowledge work, customer-service workflows and terminal-based software tasks. It estimates a cost of $0.84 per task on its tests. The model has a 500,000-token context window and standard API pricing of $2 per million input tokens and $6 per million output tokens. Requests with prompts of 200,000 tokens or more cost more.
The model is available through Grok Build, Cursor and partners including OpenRouter, Vercel and Cloudflare. SpaceXAI’s combined launch strategy is clear: Grok 4.6 supplies the model, while Grok Bot provides a consumer-friendly interface for deploying it as a continuing digital worker.
For businesses, that proposition may be attractive if agents can reliably complete tasks with limited supervision. But buyers in regulated or customer-facing sectors may also weigh the Grok brand’s past safety controversies and ongoing regulatory scrutiny around earlier products when assessing the platform.
Sources
- SpaceXAI rolls out Grok Bot AI agent app in beta on Mac, iOS, Windows, and Linux, initially for SuperGrok Heavy, Cursor Ultra, and Cursor Teams Premium users – 9to5Mac
- SpaceXAI’s Grok Bot turns agents into persistent digital coworkers that can operate your apps for $120-per-month – VentureBeat
- SpaceXAI debuts Grok 4.6, overtaking Kimi K3’s performance and matching GPT-5.6 Sol for world’s third best on Artificial Analysis – VentureBeat
- SpaceXAI’s Grok 4.6 Scores 61 on the Artificial Analysis Intelligence Index – Artificial Analysis
Stay up to date
AI for content creation: the latest tools, tips and trends. Every two weeks in your inbox:
