Runtime vs Grok Bot

Grok Bot and Runtime at a glance

Grok Bot is xAI's always-on AI teammate product, launched in beta in August 2026. Each person's bots share a persistent cloud computer with a browser, filesystem, and terminal, so work keeps going when your laptop is closed. Bots sign into your apps and websites, including ones with no clean API, and you can show a bot a workflow once and have it save the steps as a routine to rerun on a schedule. You can run several bots in parallel, with a coordinator bot managing specialists. Team Bots, in public beta since late September, share files, plugins such as Salesforce, Notion, and GitHub, credentials, and memories across a team, and can be called from Slack. It is for individuals and teams who want a capable helper for everyday knowledge work, inside a subscription many of them already pay for.

Runtime is the AI agent harness for payment and fintech teams. Payment ops, risk, compliance, finance, underwriting, onboarding, and support build agents from their SOPs. Each agent works on its own isolated computer in your cloud, reaches your ledger, processor, and bank portals through APIs, databases, MCP servers, and a browser, and stops for a person before anything moves money. It is for operations leaders who want many governed agents across the org, not one helper per person.

Good products, different jobs.

How they differ

One person's helper vs a team's operation

Grok Bot is organized around the person: your bots, your computer, your conversations, which stay private even on a shared Team Bot. That suits inbox triage, account research, recruiting coordination, and the steady stream of small tasks one person carries.

Payment work rarely belongs to one person. A single stuck payment touches four teams: support gets the ticket, payment ops traces it, finance sees a recon break, and risk or compliance may review it. Today each team investigates it separately in different tools. On Runtime it is one investigation with one record, and every run adds to a shared memory of resolved tickets, fraud caught, and reconciliation edge cases.

Where the agent computer runs and who holds the data

xAI is clear that Grok Bot runs only on Cursor-hosted cloud computers in the United States, with no on-premises or custom image option. The enterprise controls are real: per-user isolation, network allowlists, Auto Review before shell commands and writes, audit logs, ISO 27001 and ISO 42001 certification, and Zero Data Retention with model providers.

Runtime starts from a different assumption: the data cannot leave your environment. Agent computers run in your AWS, GCP, or Azure account, or you self-host the whole platform with Helm. You can bring your own sandbox provider. Credentials are masked, and card numbers and SSNs are stripped from prompts and logs before a model sees them.

Models

Grok Bot's documentation does not publish a model picker or a way to bring your own model. Runtime lets you use any model, including open-weight models served in your own cloud for PCI and PII work, and routes across harnesses (Claude Code, Codex, OpenCode) and models with fallbacks when a provider goes down.

Guardrails built for money movement

Grok Bot asks for approval on consequential actions and lets you approve drafts before they are sent. That is a sensible general pattern. Runtime's guardrails are specific to payments: agents start read-only, and approvals sit in front of releasing a payout, applying a reserve, booking a ledger entry, or replying to a sponsor bank. Every run is recorded end to end, from trigger to each query, tool call, approval, cost, and result, and the record your sponsor bank asks for stays with you.

Learned routines vs code you own

Learning by watching is one of Grok Bot's best features: show it once, correct it, and it repeats the routine. Runtime agents can go a step further once a process is solved and turn it into a script that runs deterministically as a template. That code lives in your repos, so engineers can review and own it if it becomes mission-critical.

Where Grok Bot is stronger

• Speed to value. Self-serve setup in minutes, with no implementation project.
• Learn by demonstration. Show a workflow once and it saves the steps as a routine.
• Bundled access. Included in SuperGrok and Cursor plans that many people already have.
• Consumer polish. Desktop and iOS apps, voice memos, and a thread you can pick up anywhere.
• Breadth of everyday tasks. Sales research, inbox work, office ops, and bug fixes for one person or a small team.

Pricing

Grok Bot is included with SuperGrok, SuperGrok Plus, SuperGrok Heavy, Cursor Pro, Pro+, Ultra, and Cursor Teams plans. xAI says it comes with its own usage, separate from those plans, and usage resets weekly. Enterprise access is sold separately, and enterprise pricing is not published.

Runtime has a Free tier ($0, one session), Teams from $99 per seat per month, and Enterprise with custom pricing and self-hosting. Most teams measure it against the work of analysts they were about to hire. Rain replaced $250k in vendor spend with Runtime.

Which should you choose

Choose Grok Bot if

• You want personal or small-team helpers for research, inbox, and app tasks
• Your team already pays for SuperGrok or Cursor
• A hosted cloud computer is acceptable for the data involved
• Learning a routine by demonstration covers most of what you need

Choose Runtime if

• Agents need to work card data, PII, ledgers, or payouts
• Agent computers and models have to run in your cloud
• You need approvals before money moves and a record your sponsor bank will accept
• Several teams touch the same transaction and should share one investigation
• You want an engineer who has built payment infrastructure to get the first agents live

Many teams will use both: Grok Bot for personal productivity, Runtime for the operations that touch money.

Frequently asked questions

What is Grok Bot?

Grok Bot is xAI's product for always-on AI agents, launched in beta on August 11, 2026. Bots share a persistent cloud computer with a browser, filesystem, and terminal, sign into your apps, can learn a workflow by watching you once, and can rerun it on a schedule.

Can Grok Bot run in my own cloud?

No. xAI's documentation says Grok Bot runs only on Cursor-hosted cloud computers in the United States, and on-premises or custom image deployments are not supported. Runtime runs agent computers in your AWS, GCP, or Azure account, or fully self-hosted.

Is Grok Bot a good fit for payment operations?

It can help individuals with research, inbox work, and app tasks. For work on card data, PII, ledgers, and payouts, most fintech teams need agents in their own environment, model choice for PCI and PII work, approvals before money moves, and a record a sponsor bank will accept. That is the job Runtime is built for.

How much does Grok Bot cost?

Grok Bot is included with SuperGrok, SuperGrok Plus, SuperGrok Heavy, Cursor Pro, Pro+, Ultra, and Cursor Teams plans, with its own usage separate from those plans. Enterprise access is sold separately. Runtime has a free tier, Teams from $99 per seat per month, and custom Enterprise pricing.