According to VentureBeat, SpaceXAI launched an early beta of Grok Bot on August 11, 2026 — an AI agent built to keep executing work inside a user's tools rather than just answering prompts. Grok Bot ships with Cursor Premium Teams at $120 per seat monthly, is bundled into the $200 Cursor Ultra plan, and reaches existing $300 SuperGrok Heavy subscribers too, VentureBeat and iThome report.
What Grok Bot Is: AI Agents as Persistent Digital Coworkers
SpaceXAI began launching an early beta version of Grok Bot on August 11, 2026, describing it as "a new agent designed to move AI assistants beyond answering prompts and toward continuously executing work," according to VentureBeat (E1). Taiwan's iThome corroborates the timing, reporting that "SpaceXAI周一(8/11)宣布自主代理人Grok Bot" (SpaceXAI announced the autonomous agent Grok Bot on Monday, August 11), claiming it can work "like a human," run long sessions, and coordinate across multiple agents (E13).
In its own announcement, SpaceXAI framed the product this way, per VentureBeat: "'Bots are AI teammates that do real work for you,' the company said in announcing the product. 'They sign in to your tools, use them just like you do, and come back with finished work'" (E7). That framing sits on top of a corporate restructuring: VentureBeat notes that "SpaceX acquired Cursor for $60 billion back in June" (E6) — the deal that put Cursor's coding-agent product inside the SpaceXAI/Grok family and set up Grok Bot's release eight months later.
Three Subscription Tiers, One New Feature
Grok Bot access is distributed across three existing subscription products rather than sold as a standalone plan. VentureBeat lays out the tiers directly:
| Plan | Price | What it includes (per VentureBeat) |
|---|
| Cursor Premium Teams | $120 per seat/month | Centralized billing and settings, a team marketplace for skills and plugins, shared usage analytics, SAML/OIDC single sign-on (E2) |
| Cursor Ultra (individual) | $200/month | Grok Bot included for individual users (E3) |
| SuperGrok Heavy | $300/month | Existing subscribers also receive Grok Bot access (E4) |
The $120 team seat is the headline entry point, but Grok Bot is not a separate purchase — it rides on whichever of these three plans a user or organization already holds (E2, E3, E4).
Rollout Timeline and Platform Availability
VentureBeat reports the beta "arrives for macOS, Windows, Linux and iOS, with Android listed as coming soon" (E5). iThome's account of the same announcement is more specific about what is downloadable right now: "現在Grok Bot以Beta推出Windows及iOS版,可供用戶下載試用" — Grok Bot's beta is currently available to download and try only on Windows and iOS (E14). iThome adds that enterprise access is gated behind a waitlist: "企業用戶則需先在候補清單註冊等待未來開放" — enterprise users must register on a waitlist and wait for a future opening (E15).
Read together, the two reports describe a staged rollout: VentureBeat's four-platform, Android-pending roadmap (E5) describes where Grok Bot is headed, while iThome's Windows/iOS-only download and enterprise-waitlist detail (E14, E15) describes what is actually accessible at the August 11 beta launch (E1).
Under the Hood: Cloud Infrastructure That Lets Grok Bot Keep Working Offline
iThome reports that "Grok Bot在雲端擁有獨立的運算資源,包括雲端VM、瀏覽器、檔案系統及終端,能在用戶關閉電腦或離線時持續執行任務,執行完再回覆用戶" — Grok Bot has independent compute resources in the cloud, including a cloud VM, browser, file system and terminal, letting it keep running tasks after a user closes their computer or goes offline, then reports back once finished (E17).
That architecture is what a SpaceXAI product staffer, Roman, points to as the difference between Grok Bot and a chat-based assistant, per VentureBeat: "'Grok Bot can finish the swing, because the work lands where a human would put it, in the actual tool'" (E11). The cloud VM, browser, filesystem and terminal access (E17) is the mechanism behind that claim — the agent operates inside the same tools a person would use, rather than returning a suggestion for a human to act on (E11).
Mixed Reception: Praise, Router Complaints, and a Model-Identity Dispute
Early outside reaction, as compiled by VentureBeat, is largely positive but not uniform. Lenny Rachitsky called it a standout: "'I haven't been this excited about a new AI product in a while. It's like OpenClaw, but super easy, reliable, and less scary to use. I think this will be a huge new product line for Cursor/Grok/SpaceX'" (E8). AI entrepreneur Matt Shumer described the scope of the product on X: "'The best way I can describe it is an agent for everything, not just code'" (E9).
Shumer also flagged a rough edge: VentureBeat reports that "the model router 'wasn't great' during his testing, although he said he was subsequently told it had improved" (E10). That router complaint connects to a broader disclosure gap VentureBeat identifies: "SpaceXAI's expanded announcement still does not identify which underlying models the router uses, nor does it document a mechanism for users to select, pin or switch to a particular xAI or third-party model" (E12).
iThome's reporting directly contradicts that point. It states that "Grok Bot底層是SpaceXAI和Cursor聯合打造的Grok 4.5" — Grok Bot's underlying model is Grok 4.5, jointly built by SpaceXAI and Cursor (E16). The two outlets' claims on this specific point cannot both be read as consistent: VentureBeat reports the model identity as undisclosed (E12), while iThome names it as Grok 4.5 (E16).
Inside SpaceXAI: Multi-Agent Workflows Already in Production
iThome reports that SpaceXAI is already running Grok Bot internally across multiple functions: "SpaceXAI內部已用它來執行多步驟協同任務,像是業務Bot以電話轉錄記錄更新CRM系統、營運Bot協助新入職員工、處理以Gmail寄來的發票、工程Bot可重建臭蟲、填具支援工單並將修補程式交給除錯Bot" — a sales bot updates the CRM system using call transcripts, an operations bot helps onboard new employees and processes invoices received via Gmail, and an engineering bot reproduces bugs, files support tickets, and hands patches off to a debugging bot (E18).
This matches the multi-agent coordination capability iThome cited in the launch announcement itself — that Grok Bot supports "多代理人協同" (multi-agent coordination) alongside long-running sessions (E13). The internal examples (E18) function as a concrete illustration of what that coordination claim (E13) looks like in practice inside SpaceXAI's own operations.
What This Means
The evidence points to a product being released in stages that don't fully match each other yet. The pricing is settled and specific — $120 per seat for teams, $200 for individual Cursor Ultra users, $300 for existing SuperGrok Heavy subscribers (E2, E3, E4) — but platform access is not: Windows and iOS are downloadable now, macOS and Linux are named as targets, Android is pending, and enterprise users sit on a waitlist (E5, E14, E15). The most direct tension in the record concerns what is actually running under the product's hood: VentureBeat reports SpaceXAI has not disclosed which models its router uses (E12), while iThome names the model as Grok 4.5 (E16) — a gap that lines up with Shumer's own complaint that the router itself "wasn't great" during testing (E10), even as he and Rachitsky praised the product's broader scope (E8, E9). None of the evidence gathered here resolves that discrepancy; it remains an open question between the two reports.