Its own cloud computer
Each bot runs on a persistent VM with a browser, filesystem, and terminal. Files and signed-in sessions stay on that machine — not on your laptop.
Computer & appsCursor · Grok Bot · the pipeline
I’ll set up Cursor — and show you what Grok Bot actually does.
Rules, skills, Git, and MCP in the editor. Cloud agents for PRs. Grok Bot as a standing teammate on its own cloud computer.
September 2026 cohort: Sep 8, 15, 22, 29 · 3–4pm CST · $249/seat See the weeks →
Minutes
Bug report → Cursor → push → live on client work
Always-on
Grok Bot keeps a cloud computer — work continues after you close the laptop
5 clients
Todoist ticket → Cursor cloud agent → PR
Grok Bot
What it can do
A standing teammate on a persistent cloud computer — not a one-off chat. Joint SpaceXAI + Cursor product. Sign in with your Cursor account.
Works in the real tools
Bots sign into the apps you already use. Connectors/MCP when they exist; computer use (clicks, typing, browser) when they don’t — so the job lands in the product, not a draft.
OverviewPlugins & MCP
Settings → Plugins. Installed connectors are account-wide. Type @ to attach a connector. Local stdio / localhost MCP is not reachable — the bot lives in the cloud.
Connect pluginsSkills, routines, schedules
Show it a path once. It saves a routine and re-runs on demand or on a schedule. Type / for saved skills. No workflow builder required.
Skills & routinesBots that coordinate
All of your bots share one account-scoped computer (separate screens). They message each other, share context, and pass ownership — treat a login as available to every bot.
How bots share a computerDesktop + iOS
Same thread on desktop and iPhone. Hand off a job, shut the machine, pick it up from your phone. Approvals come back when something actually needs you.
Grok Bot on mobileApprovals & secrets
Bots act on real accounts. You sign in on the cloud computer (the bot doesn’t see the password). Never paste API keys into chat — store secrets the supported way.
Approvals, security, privacyUsage is separate
Grok Bot has its own usage pool. Handing work to a bot does not burn your existing Cursor or Grok plan quota. Plans and billing live in Cursor Help.
Plans & billingPick the right surface
Editor, cloud agent, or Grok Bot
Cursor Agent (editor)
Interactive in the IDE. Rules, MEMORY.md, skills, and local MCP. Best for the file you’re in right now.
DocsCloud Agent
Isolated VM per run. Clones the repo, builds, tests, opens a PR. Best for a bounded ticket with a git artifact.
DocsGrok Bot
Standing teammate on a persistent cloud computer. Research, routines, inboxes, apps without a clean API. Not a substitute for a PR workflow.
DocsSeptember 2026 · $249/seat
Cursor is the delivery loop. The cohort is the sales system.
4 weeks · Sep 8, 15, 22, 29 · 3–4pm CST
The operating system
How I actually run Cursor
VS Code underneath — same extensions, same keybindings. The difference is the agent is wired to the repo, not a side chat.
MEMORY.md + rules
Every repo gets durable project memory and rules so the agent follows how I actually ship — not a generic coding style.
New chat per task
Keep context under 50%. Fresh thread when the job changes. The agent stays sharp instead of drowning in leftover files.
Skills and MCP
Repeatable jobs become skills. External tools come in through MCP. The agent runs your stack, not a chat tab on the side.
Ship like production
Secrets in env only. Audit RLS. No preview-branch chaos. Same git flow you already know — VS Code underneath.
The pipeline
Ticket in. PR out.
This is the loop I run for client work — and the part most people never see when they “just try Cursor.”
- 1
Ticket with the real context
Task in Todoist with comments, images, and the spec. Move it to In Progress.
- 2
Cloud agent opens the PR
A Cursor cloud agent picks it up, branches from the title, implements, and opens a PR for review.
- 3
Follow up in the PR
Need changes? Comment @cursor. It handles the follow-ups. Time goes to roadmap, not babysitting the editor.
Sprinkled in: the cohort
Four weeks if you want the whole agency loop
Setup training is Cursor. The cohort is offers, scoping, delivery, then the local LLM + Courier lab. Week 3 is where this editor workflow shows up.
Week 1 · Sep 8
Offers & positioning
Clarify who you sell to, what you sell, and the offer language that gets replies — not generic AI services.
Week 2 · Sep 15
Scoping & the sales system
Turn interest into scoped work: discovery, pricing posture, and the LinkedIn + inbound loop I use for agency pipeline.
Week 3 · Sep 22
Delivery stack & growth
Ship with a repeatable build loop — tooling, handoff, and how you grow beyond one-off projects.
Week 4 · Sep 29
Local LLM + Courier lab
Run private inference on your stack and productize flat-rate AI for clients who hate token surprise bills.
Setup training
Leave with a Cursor workflow you can repeat.
Get your development environment fully set up and learn how to ship. I'll help you configure GitHub, Git, and Cursor, set up your environment with rules and skills, and get you productive. Choose a 1:1 session or join a group session.
- GitHub and Git configuration
- Cursor setup and optimization
- Environment setup (rules, skills, full workflow)
- Hands-on training to get you productive
This is for you if
- You're switching off ChatGPT copy-paste and want Cursor set up like a delivery system
- You want GitHub, rules, skills, and MCP wired so the agent stops guessing
- You want Grok Bot explained — what it is vs cloud agents, and how to hand work off
- You'd rather get productive in one session than spend a week configuring
Not this if
- You already have a Cursor workflow you're happy with — just download it
- You want a done-for-you build (that's /build)
- You want the full sales + delivery playbook — that's the cohort
Next step
Want the full sales + delivery stack — not just the editor?
Setup training gets Cursor working. The cohort is the playbook I use to sell and ship AI work, plus the Courier / local LLM lab.
See the cohort