What Grok Bot Is, and What It Is Not
Grok Bot is xAI's persistent AI agent product, launched in beta on 11 August 2026, and it is a different product from the free @grok chatbot that answers replies on X. Every useful Grok Bot tutorial has to start there, because the naming overlap sends a lot of people looking for the wrong thing.
The chatbot answers a question and forgets you. Grok Bot is marketed as an "AI teammate". Each agent gets its own dedicated cloud computer, signs into tools and websites, learns workflows by demonstration, retains memory across sessions, coordinates with other agents, and runs 24/7 without supervision.
That changes the question you should be asking. With a chatbot you ask what it can answer. With a persistent agent you ask what you are willing to hand over and what you are keeping. A first week spent answering that is worth more than a first week spent prompt engineering.
Grok 4.6 shipped one day later, on 12 August 2026. Two releases a day apart are easy to blur together, and they are separate things: one is a model, the other is the agent product this tutorial is about.
Grok Bot itself has no free tier and no credit program, which makes the rest of your stack the only place left to recover budget. AI Perks tracks $7.7M in credits across 194 companies for exactly those layers.
Before Day 1: Check Whether You Can Actually Access It
Grok Bot has no public API and no free tier. It is subscription only, and the sources disagree about which subscription you need.
Per xAI's own page, Grok Bot is gated behind SuperGrok, SuperGrok Plus and SuperGrok Heavy, and on the Cursor side behind Pro, Pro+, Ultra, Teams Standard and Teams Premium. Desktop and iOS at launch, Android described as planned, enterprise access by waitlist.
Then there is the discrepancy worth knowing about. Secondary coverage of the launch narrows real access to the top tiers only, quoting figures around $300/month for SuperGrok Heavy, $200/month for Cursor Ultra and $120 per seat per month for Cursor Teams Premium. That conflicts with the broader tier list on the official page.
Treat xAI's own page as authoritative and then check what your specific account shows. Do not budget off a number in an article, including this one.
| Grok Bot access question | Straight answer |
|---|---|
| Free tier | None |
| Public API | None |
| Startup credit program | None |
| Trial | None published |
| Platforms at launch | Desktop and iOS |
| Android | Described as planned |
| Enterprise | Waitlist |
If you are wondering why a Cursor subscription unlocks an xAI product at all: SpaceX acquired xAI in an all-stock merger announced 2 February 2026, and SpaceX/xAI acquired Anysphere, Cursor's parent company, in June 2026. The bundling follows the ownership.
Days 1 and 2: What to Delegate First
Delegate work that is repetitive, verifiable, and cheap to get wrong. Those three conditions together are the whole filter, and a task that fails any one of them belongs in next week, not this one.
Strong first-week candidates:
- Monitoring and triage. Watching a queue, an inbox, a dashboard or a channel and flagging what matters. The agent runs 24/7, which is exactly the shape of this work.
- Cross-tool data gathering. Pulling the same fields out of three systems that do not talk to each other.
- Recurring reporting. Anything you rebuild on the same cadence from the same sources.
- Mechanical first drafts. Structure and assembly, not final wording.
- Routine tool operation you already do by hand. Because Grok Bot learns workflows by demonstration, the strongest day-one move is to perform a task once while it watches rather than write a long prompt describing it.
That last point is the part most tutorials skip. Demonstration carries the details you would forget to write down: which tab, which filter, which edge case you always handle silently.
Days 3 and 4: What to Keep
Keep anything where being wrong is expensive, irreversible, or invisible. Invisible is the one people underweight, and it is the one that hurts.
Keep on your side of the line:
- Anything that spends money. Purchases, plan upgrades, ad budgets, API usage that bills by token.
- Anything sent externally in your name. Publishing, customer replies, contracts, outreach.
- Credential and access decisions. An agent that signs into tools is exactly why this matters.
- Final judgment calls. Hiring, pricing, positioning, anything where the reasoning matters more than the output.
- Work you have no cheap way to check. If verifying the result costs as much as doing it, delegation buys you nothing.
One honest note from the early beta: a reviewer found that the automatic model router "wasn't great". That is a UX critique rather than a scandal, but it is a good argument for keeping a human read on output during week one instead of trusting the default routing blindly.
The First-Week Schedule
Run the week as an escalation, not a big bang: observe, demonstrate, supervise, then release one narrow task to run unattended.
| Day | Focus | Delegate | Keep |
|---|---|---|---|
| 1 | Setup and disambiguation | Nothing yet | All of it |
| 2 | Demonstrate one workflow | One read-only task | Every write action |
| 3 | Supervised runs | Monitoring and triage | Sending and spending |
| 4 | Add a second task | Cross-tool data pulls | Final wording |
| 5 | Review the logs | Recurring reports | Judgment calls |
| 6 | Narrow the scope | One unattended task | Credentials and access |
| 7 | Decide what week two gets | Whatever passed review | Everything that did not |
The point of day 7 is subtraction. Most teams end week one having handed over too much, and the fix is to take two things back rather than add a third.
If week two means adding tools around the agent, check which of them run credit programs before you pay for any of them at getaiperks.com.
What the Agent Stack Around It Costs
Grok Bot itself has no credits to chase, but almost every other layer of an agent stack does, and that is where the budget is actually recoverable.
A persistent agent does not run alone. It calls models, stores state, runs code somewhere, sends email, queries data. Those layers have startup credit programs even when the agent product does not.
AI Perks tracks $7.7M in credits across 194 companies, which is the practical answer to a Grok Bot subscription you cannot discount: stop paying full price for the eight other tools in the stack.
If You Cannot Access Grok Bot: Use the Grok API Instead
The Grok API is a different product from Grok Bot, with a different access model, and it is what you would use to build your own Discord or Telegram bot. People searching for a Grok bot tutorial usually want one of these two things, and the API is the one with a public sign-up path rather than a subscription gate.
Three things that are useful to know before you start:
- A new-account API credit of $25/month has been reported as current for the Grok API.
- An optional data sharing arrangement offers a recurring $150/month credit in exchange for sharing API request metadata and outputs. Whether that trade is acceptable depends entirely on what your bot handles.
- Token pricing varies by model and generation, and published figures disagree badly across sources. Check xAI's own console for current rates rather than trusting any number you read in an article.
xAI has no dedicated startup credit program comparable to the ones run by other major labs. If you were counting on one to fund a build, plan around its absence. The credits you can actually stack sit elsewhere in the stack, and AI Perks is where to find which ones are live right now.
Frequently Asked Questions
Is Grok Bot the same as the @grok chatbot on X?
No. The @grok chatbot is the free conversational assistant that replies on X. Grok Bot is xAI's separate persistent agent product, launched in beta on 11 August 2026, sold by subscription, with its own dedicated cloud computer per agent and memory that carries across sessions.
Does Grok Bot have a free trial or an API?
Neither. Grok Bot has no public API and no free tier, and xAI has published no trial. Access comes through SuperGrok or Cursor subscription tiers only. If you need programmatic access, the separate Grok API is the product to use, and AI Perks tracks credits for the rest of your stack.
Which subscription do I need for Grok Bot?
xAI's own page lists SuperGrok, SuperGrok Plus, SuperGrok Heavy, and Cursor Pro, Pro+, Ultra, Teams Standard and Teams Premium. Secondary coverage narrows this to the top tiers only. The sources conflict, so treat xAI's page as authoritative and confirm against what your own account actually shows.
What should I delegate to Grok Bot first?
Repetitive, verifiable, low-cost-of-error work: monitoring and triage, cross-tool data gathering, recurring reports, mechanical first drafts. Because it learns workflows by demonstration, perform the task once while it watches instead of writing a long prompt. Keep spending, sending and credential decisions on your side.
Are there free credits for Grok Bot?
No. Grok Bot has no credit program, no free tier and no startup offer. The separate Grok API has a reported $25/month new-account credit and an optional data sharing credit. For the layers that do run credit programs, AI Perks covers $7.7M across 194 companies.
How does Grok Bot compare to other persistent agents?
Its direct competitors in the persistent-agent category are ChatGPT Agent and Claude Cowork. xAI has not published a feature comparison against either, so the useful move is to read each vendor's own product page on the three things that decide a first week: which platforms are supported, which subscription unlocks the agent, and how you hand it a workflow.
Delegate the busywork, keep the judgment, and stop paying list price for the rest of the stack at getaiperks.com.