Grok Bot in Cursor: a chat-window teammate for the sub-tasks you keep punting

What it is

Grok Bot is xAI's new "AI teammate" product, shipped jointly with Cursor and reachable through an iMessage-style chat window on desktop and mobile. xAI has launched Grok Bot, a beta product it describes as a team of always-on AI agents that get their own cloud computer, sign into a customer's existing tools, and finish multi-step jobs without being supervised. For developers, the pitch is that each bot is a persistent agent you can DM a task to — not a chat you re-prompt every session.

The next-work-session test

Concrete scenario: you're mid-flow in Cursor debugging an SDK integration and you notice your eval harness is failing on three fixtures. Instead of context-switching to fix them yourself, you open the Grok Bot chat, DM a bot named eval-janitor, and hand it: "reproduce these three failing cases against staging, capture logs, and file a bug with the trace." The bot goes off on its own cloud computer, logs into your bug tracker, and comes back with a ticket while you keep coding.

What changes: the "little chore" tax on the agent loop — reproducing bugs, updating a fixture in a Notion doc, kicking off a manual eval — moves from your terminal to a chat message. The idea behind Grok Bot: users are meant to hand work to a bot the way they would to a colleague – via chat, without having to configure workflows, automations or routines first.

Pricing

Verified. There is no standalone Grok Bot price — access is gated behind three existing subscriptions. Grok Bot's pricing page lists two self-serve plans: Cursor Ultra at $200 per month and Cursor Premium Teams at $120 per seat per month, both billed monthly. It is also included if you already pay for SuperGrok Heavy or Cursor Ultra. SuperGrok Heavy is priced at $300/month per xAI's tier structure, though xAI does not list the price on its public pricing page, where individual plans stop at SuperGrok Plus. Free, SuperGrok at $30, and SuperGrok Plus at $100 do not include Grok Bot as of August 12, 2026. Cursor Pro and Pro+ are also out — you need Ultra or a Premium Teams seat.

Full plan mechanics are on the Cursor Grok Bot plans and billing page.

What we'd actually use it for

Narrower than the marketing. In practice, we'd point one bot at a bounded, low-stakes lane: eval reruns, dependency update PRs, scraping release notes from an SDK vendor's changelog, or turning a Slack thread into a Linear issue with the repro steps attached. If your bottleneck is writing correct software, the existing harness comparison still governs. If your bottleneck is the forty minutes a day you spend logging into SaaS tools to move small pieces of state around, that is the gap this targets.

Translation for developers: this is not going to out-code your Cursor Composer session. It's for the SaaS-tab work around the code.

Limits

It's early beta, and the honest caveats are worth reading before you connect anything sensitive:

• Skill capture is narrow. Ask a Bot to watch you do something once in the computer view, and it writes a skill from the recording. It's capped at ten minutes, browser-only, and doesn't capture microphone audio. Fine for UI flows, useless for anything terminal-based.
• No published reliability data. Plan-by-plan differences in Bot limits and concurrency. Enterprise launch timing and pricing. Default approval thresholds, failure recovery, and practical beta reliability are all still unclear.
• Not a coding agent replacement. It's a separate product line from xAI's terminal coder. Grok Bot is a distinct product line — an "AI teammate" app for desktop and iPhone, not a terminal coding agent and not a prompt-to-app builder.
• Blast radius is real. If you already qualify through Cursor Ultra, Cursor Premium Teams or SuperGrok Heavy, TRY Grok Bot on one isolated, reversible, low-stakes workflow. Do not give the early beta broad access to production systems, financial accounts, regulated data or an executive's primary browser profile. Give it a scoped OAuth token, not your prod admin login.

Try it if

• You already pay for Cursor Ultra, Premium Teams, or SuperGrok Heavy — the marginal cost is zero.
• You lose real time each week to SaaS-tab chores that sit around your agent loop (bug filing, changelog scraping, eval kickoffs).
• You want to experiment with delegated sub-agents that persist between Cursor sessions.

Skip it if

• You're on Cursor Pro or Pro+ — you're not eligible, and paying $200/month just for chat-window agents is a hard sell.
• Your bottleneck is code quality inside the editor, not the peripheral SaaS work. Your existing Cursor agent still wins there.
• You'd need to hand it credentials to production, finance, or regulated systems. Wait for the audit story to mature.
• You need documented eval numbers or SLAs before adopting. There aren't any yet.