Start Your Business with Grok Bot in 60 Minutes: the Step by Step Guide

Right now anyone can build a company that generates $100k per month without hiring a single person. Not because the work disappeared. Because the work moved to AI.

xAI just launched Grok Bot and described it with one phrase that changes everything: AI teammates you can give real work to. Not a chatbot. Not an assistant. A teammate with a computer, tools, memory and the ability to complete jobs autonomously while you sleep.

Moonshot built something that goes one step further. Kimi K3 with Agent Swarm doesn't just give you one AI employee. It gives you an entire AI organization that one founder can direct from a single seat.

The math is simple. 56 clients at $1,500 per month equals $84,000 monthly recurring revenue. 100 clients at $999 per month equals $99,900. One founder. Zero employees. AI handles the work.

Here's the exact system.

What Grok Bot actually introduced

Most people saw the announcement and thought impressive demo. That's the wrong frame.

xAI didn't launch a better assistant. They launched a new software category.

Old category:
Tool → you use it → output

New category:
AI Employee → you delegate → it works → output

Grok Bot has a persistent cloud computer. It has a browser, filesystem, terminal, apps and accounts. It can log into web applications, work with files, execute commands and move between systems until a job is done.

xAI's own teams started using Bots internally before the public launch for sales outbound, marketing campaigns, office operations and bug fixes. The company that built the AI employee started using AI employees itself.

And xAI was very specific about what makes a good Bot. Not a generic helper. A worker with a defined job.

WRONG:
General Helper Bot

RIGHT:
Talent Scout Bot
Expense Manager Bot
Account Health Bot
Bug Reproduction Bot

Jobs. Not prompts. Not chats. Jobs with clear scope, tools, approval boundaries and recurring schedules.

What Kimi K3 adds to this picture

github.com/MoonshotAI/Kimi-K3

Grok Bot gives one AI employee a computer. Kimi K3 gives you the ability to build an entire AI organization.

Moonshot built K3 specifically for long-horizon coding, tool orchestration and agentic knowledge work. Not chat. Work.

Kimi K3 alone:
Brain → answers questions

Kimi K3 with full stack:
Brain
+ Hands (Kimi Code)
+ Memory (Skills + Context Graph)
+ Tools (MCP)
+ Workers (Agent Swarm)
+ Control (Approvals)
= AI Company

The key number from Moonshot's documentation: Agent Swarm supports up to 300 subagent instances executing over 4,000 tool calls in a single task with up to 4.5x faster execution compared to sequential single-agent approaches in large search scenarios. These are Moonshot's own benchmarks not independent verification but the architecture is real and the primitives are available right now.

One founder directing 300 AI workers on a single task is not science fiction. It's a documented feature.

The org chart of an AI company

Here is what a real AI company built on this stack looks like:

                      FOUNDER
                       │
               sets goals, approves
               high-stakes decisions
                       │
                    KIMI K3
                  AI Executive
                       │
          ┌────────────┼────────────┐
          │            │            │
     PERMANENT     RESEARCH     ELASTIC
       TEAM          DEPT       WORKERS
          │            │            │
    Grok Bot      Main Agent    K3 Swarm
    pattern       + Subagents   up to 300
          │            │            │
    Sales Bot    Research     Temporary
    Ops Bot      Analysis     Researchers
    Dev Bot      Synthesis    Analysts
    Finance Bot  Reports      Coders
          │            │            │
          └────────────┼────────────┘
                       │
                     SKILLS
                (encoded expertise)
                       │
                      MCP
                 (company tools)
                       │
        ┌──────────────┼──────────────┐
        ▼              ▼              ▼
       CRM           GitHub        Database
       Email         Browser       Payments
                       │
                    ROUTINES
               (24/7 work schedule)
                       │
                  APPROVAL GATE
                       │
                    FOUNDER

Two types of AI workers. Permanent team that handles recurring jobs with institutional memory. Elastic workforce that scales up for large research tasks and scales back down when done.

The founder sits above both. Not doing the work. Approving the decisions that matter.

Layer 1 - Grok Bot pattern: permanent AI employees

docs.x.ai/grok-bot/bots

A Grok Bot is a durable AI teammate with a name, a job, its own conversation and working context that develops over time.

The key word is durable. Not a session. Not a prompt. A persistent worker that accumulates context about its domain the same way a human employee builds knowledge over months on the job.

A Bot should be created when a role has distinct goals, tools, working style and approval boundaries. One Bot per job function.

Sales Bot
└── researches leads
└── prepares call briefings
└── drafts follow-up emails
└── updates CRM after every interaction
└── flags deals above €50k for founder review

Ops Bot
└── monitors suppliers
└── tracks delivery timelines
└── flags anomalies in data
└── prepares weekly operations report

Dev Bot
└── reviews PRs before humans see them
└── runs tests on schedule
└── drafts documentation after code merges
└── flags security issues for review

Each Bot can have up to 50 recurring routines. Background routines run when the laptop is closed. Event-driven routines wake up when something happens.

New lead comes in → Sales Bot wakes up
New GitHub issue → Dev Bot wakes up
Customer complaint → Support Bot wakes up
Every Monday 8am → Reporting Bot starts
Every Friday 5pm → Finance Bot reconciles

The founder doesn't trigger any of this. The company starts its own workday.

Layer 2 - Kimi Code: giving AI hands

github.com/MoonshotAI/kimi-code

Grok Bot has a persistent computer. Kimi Code gives K3 the same capability inside your own stack.

Kimi Code can read and edit files across an entire codebase, run shell commands, search for information, fetch web pages, analyze results and independently choose the next action based on what it finds. It supports MCP, lifecycle hooks, subagents and Skills.

Without Kimi Code:
K3 tells you what should happen

With Kimi Code:
K3 does what should happen

A real example. You delegate client research to your Sales Bot. Without execution capability it writes you a list of things to look up. With Kimi Code it opens the client's website, reads their LinkedIn, searches their recent funding, pulls their job postings as a signal of growth strategy, checks if they use competitor products and writes a structured briefing directly into your CRM. You come back to a finished output.

Layer 3 - Agent Swarm: elastic AI workforce

kimi.com/en/help/agent/agent-swarm

This is where Kimi K3 goes beyond what a single Grok Bot can do.

Agent Swarm lets the main K3 agent decide how to parallelize work, which specialists to create and how to delegate. The founder describes the goal. K3 designs the organization needed to achieve it.

Founder: "Analyze 100 markets and find the best one for our SaaS."

K3 thinks: "This is parallelizable."

K3 creates:
Market 1 → Research Agent
Market 2 → Research Agent
Market 3 → Research Agent
...
Market 100 → Research Agent

All run in parallel
↓
Each returns structured findings
↓
K3 synthesizes into ranked recommendation
↓
Founder receives final report

The honest caveat from Moonshot: more agents doesn't automatically mean better results. Coordination quality matters more than agent count. K3 itself warns that subagents consume model tokens separately so simple tasks are better handled by the main agent alone.

Simple email → 1 agent
Research 100 companies → Swarm
Build a feature → planner + coder + reviewer
Check one number → no agents needed

The goal is maximum output per dollar. Not maximum agents.

Layer 4 - Skills: encoding the job

kimi.com/code/docs/en/kimi-code-cli/customization/skills.html

This is the most important layer for building a defensible business.

A Skill is a markdown file that encodes exactly how a specific job should be done. Workflow, rules, constraints, what to never do. K3 reads it automatically when the task matches.

Real Skill for a recruiting AI company:

---
name: candidate-research
description: Research and score candidates for technical roles
whenToUse: When evaluating a new candidate application
---

## Step 1 - Profile review
Read their LinkedIn, GitHub and portfolio.
Check recent contributions and project quality.
Note employment gaps and career trajectory.

## Step 2 - Skills assessment
Map their stated skills against role requirements.
Check GitHub for actual code quality not just listed skills.
Look for domain-specific experience signals.

## Step 3 - Culture signals
Read their public writing if available.
Check how they communicate in open source projects.
Note any red flags in public professional behavior.

## Step 4 - Scoring
Score 1-10 on technical fit, experience level, culture signals.
Write a two-paragraph summary for the hiring manager.
Flag any concerns that need human review.

## Rules
Never contact candidates directly.
Never score below 7 without a specific written reason.
Always flag if resume shows signs of AI generation.
Escalate any legal red flags to founder immediately.

After writing this once the AI hiring assistant follows your exact process on every candidate forever. No re-explaining. No variation. Every mistake you catch becomes a rule that prevents the next occurrence.

You don't train a new model to create an AI employee. You encode the job into a Skill.

Layer 5 - MCP: connecting AI to company infrastructure

Clients don't pay €500 per month for a smart chat interface. They pay for a system that reads their CRM, drafts their proposals, updates their tickets and follows their actual workflow.

MCP is what makes the AI employee part of a company's infrastructure rather than a tool sitting next to it.

Kimi K3
↓
MCP
↓
GitHub    reads repos, opens PRs, comments on issues
CRM       reads pipeline, updates records, creates tasks
Database  queries live business data directly
Email     drafts and sends with approval
Slack     posts updates, pings humans on escalations
Browser   researches competitors, pulls public data

Every MCP connection is a capability you're packaging into your product. Every restriction is a trust boundary you're selling to enterprise clients.

Layer 6 - Approvals: the founder stays in control

This is the layer that makes AI employees production-safe.

xAI recommends approval gates on external messages, publishing, purchases, financial transfers, deletion and production changes. Kimi Code has permission rules that determine what runs automatically versus what requires human sign-off.

AI COMPANY - WHAT RUNS AUTOMATICALLY:
Research
Analysis
Drafting
Calculation
Code writing
Test running
CRM reading
Report generation

AI COMPANY - WHAT REQUIRES FOUNDER:
Sending contracts
Deploying to production
Deleting records
Purchases above limit
Contacting clients directly
Publishing content

The founder stops being a worker and becomes a decision maker. The AI handles volume. The human handles judgment.

The businesses you can build

The opportunity is not selling access to Kimi or Grok. It's packaging AI workers into a specialist that understands one expensive job better than a generic tool ever will.

Recruiting operations for agencies:

Research Agent finds companies hiring
Candidate Agent matches against requirements
Outreach Agent drafts personalized messages
Interview Agent prepares briefing notes
CRM Agent updates records after every interaction
Founder approves shortlists and offers

Price: €1,500/month per agency
56 clients = €84,000/month
Infrastructure: ~€800/month
Margin: ~99%

Legal research for small firms:

Research Agent reviews case precedents
Document Agent summarizes contracts
Flag Agent identifies relevant clauses
Brief Agent prepares lawyer briefings
Founder reviews before anything goes to clients

Price: €2,500/month per firm
34 clients = €85,000/month
Infrastructure: ~€600/month
Margin: ~99%

E-commerce operations:

Inventory Agent monitors stock levels
Content Agent updates product descriptions
Supplier Agent responds to routine emails
Analytics Agent flags anomalies in sales data
Founder reviews weekly summary

Price: €499/month per store
200 clients = €99,800/month
Infrastructure: ~€1,200/month
Margin: ~99%

You're not selling parameters. You're selling outcomes. The Skill files encode the expertise. The MCP connections encode the workflow. The approval rules encode the trust boundaries. That combination is what competitors can't easily copy.

How to build your first AI company this week

Day 1 - pick one expensive job that businesses currently pay humans to do. Sales research, contract review, candidate screening, inventory management. The job should be repetitive, research-heavy and have clear deliverable output.

Day 2 - run the job manually with Kimi Code. Watch exactly what it does, where it gets stuck and what information it needs. This is your process documentation.

Day 3 - encode the process into a Skill file. Every step, every rule, every constraint. Test it on five real examples. Fix what breaks.

Day 4 - connect MCP to the tools the job actually needs. CRM, database, email, browser. Verify each connection works correctly.

Day 5 - set approval boundaries. Define what runs automatically and what requires human sign-off. Build in the trust layer that makes clients comfortable.

Day 6 - build the Agent SDK wrapper and turn the interactive session into a software product that clients can access.

Day 7 - show one real client the agent doing their actual job on their actual data. Not a demo with fake data. Their real workflow.

Then charge for it.

What this actually means

xAI launched Grok Bot and used three words that reframe everything: AI teammates. Not assistants. Teammates with computers, tools, memory and jobs.

Kimi K3 takes that one step further. The main agent can decide how to organize work, how many specialists to create and how to synthesize results. The founder describes the outcome. The AI designs the organization needed to achieve it.

The companies that win the next decade won't be the ones with the most employees or the most parameters. They'll be the ones that encode the most domain expertise into Skills, connect the most relevant tools through MCP and build the most appropriate approval structures for their markets.

One founder. Grok Bot pattern for permanent workers. Kimi K3 Swarm for elastic workforce. Skills for institutional knowledge. MCP for company infrastructure. Approvals for human control.

That's the AI company. It's buildable today. The primitives exist. The documentation is public.

Most founders will keep hiring humans for jobs AI can already do. A few will spend one week building the system above and start collecting $100k per month from clients who can't figure out how they're doing it so cheaply.

/ If this was useful - follow, the next one drops here first.