Grok Bot Types Guide: How to Build and Staff Specialist Bots

Published:

People search “Grok Bot types” expecting a fixed menu—Research, Writer, Publisher and so on. On SpaceXAI’s product surface, that menu does not exist. What you actually create and install are Personal Bots, Team Bots, and Templates, plus named Marketplace packs and role recipes you write yourself [1][2][3].

This guide puts how-to with Grok Bot first: six practical walkthroughs for building a clear job, installing Marketplace packs, choosing Personal vs Team vs Template, running group-chat handoffs, saving skills and routines, and staffing a newsroom Research→Writer→Reviewer pipeline with approvals. The full Marketplace catalog as fetched on 7 October 2026—with practical implementation ideas—follows as the reference body.

One clarifying point: labels such as Research, Writer or Publisher are user jobs and packs, not a closed product enum. Official docs define Personal Bot, Team Bot and Template. Role names in Guides, group-chat examples or tutorials are recipes you create or install [1][2][4].

1. How to create a Personal Bot with a clear job

A Bot is a durable teammate: a name, a job description, its own conversation, and context that compounds over time. Docs discourage a vague “General Helper”—give each Bot one clear job [2].

  1. Open Grok Bot and choose New → Create new Bot [2][5].
  2. Set a short name that you can @-mention later (for example Research or Inbox Desk).
  3. Write an operational description that covers: primary goal; sources and tools it may use; deliverable format; style; schedule if any; and a hard “never” list (for example “Never send external messages, spend money, or publish without my approval”) [2].
  4. Edit the profile (label, avatar) if useful; pin Bots you use often. Switch via the sidebar, shortcuts, or search [2].
  5. Start with read-and-prepare tasks only. Correct voice and format in chat; put lasting rules in the description, not only in ephemeral messages [2][5].
  6. Connect plugins when the job needs them; complete any human takeovers for passwords, 2FA or CAPTCHA on the shared cloud computer [5].

All personal Bots on one account share a single persistent cloud computer (browser, filesystem, terminal). That is why specialist handoffs work without re-logging in everywhere [1].

2. How to add a Marketplace template and connect tools

The Marketplace lists connectors, skills and featured/shared Bots. Adding a pack installs a copy—a template recipe—not the publisher’s computer, logins or full history [2][3].

  1. Open x.ai/bot/marketplace and browse Featured or a category strip [3].
  2. Open a pack and choose View details—read skills, plugins, memories and the description before you commit [3][5].
  3. Choose Add to Grok Bot. Shared Bots are created by other users; adding one accepts third-party bot terms [2][3].
  4. Connect the plugins the job needs (email, CRM, calendar, Slack, and so on). Finish any takeover prompts on the shared computer [5].
  5. Put lasting approval rules in the Bot description. Run a small read-and-prepare task before anything external [2].
  6. When the workflow is stable, save a skill (/), then optionally a routine—see section 5 [6].

Do not confuse Template (shareable recipe → recipient gets a copy) with Team Bot (one shared seat the whole team chats with) [7].

3. Personal vs Team vs Template — when to use which

Construct Use when How you set it up What recipients get
Personal Bot The job is yours alone; memory and logins should stay private New → Create new Bot; edit profile and description [2] Only you chat with it
Team Bot The whole team needs one shared seat (private chats per person); one owner maintains plugins, secrets, skills and files New → Create new Team Bot, or Share → Publish to Team from a personal Bot; then Publish to team; Slack optional. Teammates: New chat → Team Bots → Add [7] Shared teammate; usage counts against the person talking
Template You want to share a recipe (instructions, selected memories, skills, plugins) without handing over your computer Share → Create template → Public or Team-only link → recipient Add to Grok Bot [2][8] A copy—not your computer, logins, secrets or full history

Practical rule of thumb: start Personal for a clear job; publish a Template when others should install the same operating style; use a Team Bot when one shared seat and owner-maintained tooling matter more than each person owning a private copy [7][8].

4. How to run group chat and handoffs

Specialize first—one Bot per distinct goal, tool set and approval boundary—then glue them together [2].

  1. Create 2–6 specialist Personal Bots (for example Research, Writer, Reviewer) with distinct descriptions and “never” lists [2][4].
  2. Start a group chat with those Bots plus you. Group chats support visible handoffs; address work with @Bot or @everyone [4].
  3. Kick off with a single owner per stage. Collaboration docs exemplify user-named @Researcher → @Writer → @Reviewer in one thread—do not publish without a human [4].
  4. For async work, use Bot→Bot DMs: the receiving Bot wakes and continues; you still see the handoff [4].
  5. Move bulky context via shared computer files so specialists do not re-authenticate for every handoff [1].
  6. Optional: add a coordinator (Project Manager / “Bot Boss”) as the front door so you are not the router—common in SpaceXAI Guides [9][10].

Keep irreversible actions (send, publish, spend, production change) behind your approval forever for newsroom-style work [4][6].

5. How to save skills and routines

Once a chat workflow is reliable, turn it into reusable automation [6].

  1. Run the task manually until the steps, validation checks and approval boundaries are clear.
  2. Save a skill—a reusable how-to for that task—and invoke it later with / [6].
  3. Optionally install packaged skills from the Marketplace when they match the job [3][6].
  4. Create a routine (schedule or event trigger) that runs a workflow for one Bot. Docs note a ceiling of 50 routines per Bot [6].
  5. Use Test run carefully: test runs do real work. Prefer prepare-then-approve for anything external [6].
  6. Put failure and stale-data policy in the Bot description or skill notes so scheduled runs do not silently invent updates [2][6].

Skills capture how; routines capture when. Neither replaces the approval boundary for send, publish or spend [6].

6. Newsroom pattern: Research → Writer → Reviewer with approvals

This is a staffing pattern, not an official type list. It matches the collaboration docs’ Researcher / Writer / Reviewer example and suits editorial pipelines that must stay draft-only until a human says yes [4].

  1. Create three Personal Bots with distinct jobs [2]:
    – Research — gather sources, dated claims, quotes; never invent citations; never publish.
    – Writer — turn approved research into draft copy in house style; never send or publish.
    – Reviewer — check facts against Research notes, flag unsupported claims, return a punch list; never publish.
  2. Put the same hard rule in every description: no external send, no live publish, no spend without approval [2][4].
  3. Open a group chat with all three plus you. Assign stages with @Research, then @Writer, then @Reviewer [4].
  4. Move research notes and draft files on the shared computer so each stage starts from the same artefacts [1].
  5. Human (or a draft-only “Publisher” Bot that only creates drafts in your CMS/scheduler) performs the final publish step—you approve [4].
  6. When the pipeline is stable, save a skill for each stage and optionally a routine for recurring desks (for example a morning research digest)—still prepare-then-approve [6].

Optional one-liner on runtime: around 7 October 2026, Musk said Grok Bot would route across multiple back-end models for best outcome; that is company-reported runtime intent, not a bot-type taxonomy [11].

Marketplace catalog (7 October 2026)

Inventory below matches the Marketplace page as fetched on 7 October 2026. Overlaps across Featured and a category strip are normal. Blurbs paraphrase Marketplace descriptions; they are not a second product taxonomy. Implementation ideas under each category are practical tips for how you might run the pack—not official product documentation. Re-check the live page before you treat any pack as still Featured [3].

Roughly 25 distinct named packs appeared across sections when this catalog was fetched (counting overlaps once).

Featured / From Grok Bot Team

Pack What it does (Marketplace summary)
Outbound Prospecting (Krista Letz) Finds ICP-matching prospects on the public web and drafts first messages.
SEO & AEO Desk (Adam Tanguay) Turns keywords (or Search Console input) into content ideas and writer-ready briefs.
Haggle Bot (Daniel Gartshein) Inventories SaaS spend (e.g. Ramp/bills), builds savings evidence, drafts vendor counters.
Recruiting Coordinator (Tommy Hansen) Interview loops, interviewer prep, chase stalls—without emailing candidates on its own.
Projects Manager (Eric Zakariasson) Project/PM-style Bot in the Featured strip.
dr eggbot, Overheard, Tradbot Featured community packs with identity- or brand-specific framing—read template details before Add.

Implementation ideas (practical, not official docs): Connect CRM + email (or LinkedIn-style research via browser) for Outbound Prospecting; paste an ICP one-pager into the description and run “shortlist 10 + draft first touch—do not send.” For SEO & AEO Desk, connect Search Console if available, or paste a keyword list; first task: “five briefs with title, intent, outline, internal-link targets.” Haggle Bot: connect expense/Ramp-style tools, ask for a ranked savings memo and draft counters only. Recruiting Coordinator: ATS + calendar; first task: “map this week’s loop and prep packs—no candidate email.” Projects Manager: issue tracker + Slack; use as a front-door coordinator in a group chat with specialists [3][6][9]. For community Featured packs (dr eggbot, Overheard, Tradbot), open View details and treat identity/brand framing as optional flavour—do not grant send/spend until you trust the recipe. Approval boundary across Featured blurbs: draft and prepare; do not send, spend or sign without you [3][6].

Engineering

Pack What it does (Marketplace summary)
Researchy (Farzad) Research and fact-check desk with live web search and sourced, dated claims.
Lingxi’s Engineer Bot (Lingxi Li) Engineering supervisor: board work, launch cloud agents, watch PRs (~30 min), ask before merge.
tinkabot (Lauren Tan) Wraps APIs into Cursor/Agent plugins (MCP + skills).
QA bot (Ulysses Ng) Live-deploy acceptance checklist with pass/fail before ship.

Implementation ideas: Researchy pairs well as the Research seat in a group chat—first task: “sourced brief with dated URLs; no publish.” Lingxi’s Engineer Bot: connect issue board + GitHub/Cursor cloud agents; first task: “triage open PRs and propose merge order—ask before merge”; routine idea: periodic PR watch (~30 min cadence as the blurb suggests), still human-merge. tinkabot: point it at an API docs URL and ask for an MCP/skill scaffold; keep secrets in Bot secrets, not chat. QA bot: staging URL + acceptance checklist file on the shared computer; first task: “run checklist, return pass/fail only—no production changes.” Never unsupervised merge, deploy or production edit [3][6][13].

Sales

Pack What it does (Marketplace summary)
GTM Loop Closer (Jon Grigull) Finds dropped promises and follow-ups across meetings, email, Slack and CRM; prepares replies and tasks.
Sales Call Coach (Daniel Brill) Scores calls from transcript or recording and flags what to fix next.

(Outbound Prospecting also appears under sales-related Featured listings.)

Implementation ideas: GTM Loop Closer—connect calendar/meetings + email + Slack + CRM; first task: “list dropped promises this week with draft replies—do not send”; routine idea: Monday morning open-loop digest. Sales Call Coach—drop a transcript or recording; ask for scorecard + three coaching fixes; keep coaching notes private to the rep unless you deliberately share. Outbound Prospecting (Featured): same ICP + draft-only pattern as above. Approval: no send, no CRM mass-update, no enrol without you [3][6][12].

Marketing

Pack What it does (Marketplace summary)
Stills & Clips Desk (Matt Palmer) Stills, thumbnails and clips from footage; screenshot cleanup; caption/alt text.
Clip Bot (This Week in AI) Best moments → short captioned clips plus transcript/timestamps.
Copy Humanizer (Massimo De Luisa) Rewrites drafts to sound human; shows changes; does not invent facts.

(SEO & AEO Desk is also a marketing-facing Featured pack.)

Implementation ideas: Stills & Clips Desk / Clip Bot—drop footage or a long cut on the shared computer; first task: “three thumbnail stills + two 30–45s clips with captions/alt”; do not auto-post to social. Copy Humanizer—paste a draft; ask it to show a diff-style change list and forbid new facts. SEO & AEO Desk—feed keywords or GSC, hand briefs to a Writer Bot in group chat (section 6). Routine idea: weekly “unused keyword → brief” digests that stop at draft. Never unsupervised social publish or ad spend [3][6][9].

Design

Pack What it does (Marketplace summary)
Imogen (Kent C. Dodds) Alt text focused on important image content (accessibility).
figma bro (John Bai) Figma → build spec; audit tokens/components; brief → screens.
Critiquito (Manuel Muñoz Solera) Screenshot/Figma → ranked design critique; never edits files.
Game Art Director (Danny Limanseta) Concept → style guide, palettes, prompt sheets; sprite checks.

Implementation ideas: Imogen—batch-export images to a folder on the shared computer; first task: “alt text for each file, priority on informative content.” figma bro—connect Figma; first task: “build spec from this frame + token/component audit”; keep implementation PRs human-reviewed. Critiquito—screenshot or Figma link; ask for ranked critique only (the blurb says it never edits files—keep it that way). Game Art Director—concept note → style guide + palette + prompt sheet; sprite checks before handoff to art production. Approval: no live design-file edits or ship without you [3][6].

Personal

Pack What it does (Marketplace summary)
Credit Card Max (Trevin Chow) Which card to use; benefits and utilisation review.
Home robots (Sawyer Merritt) Control home robots (Navimow, Matic, Matter and similar) from chat.
Flora: Plant Care Log (Rich Silver) Private houseplant journal and reminders.
Pfp Bot (Shub Gaur) Grokbot-style icon from a photo; no realistic likeness posts.

Implementation ideas: Credit Card Max—list cards and rewards rules in the description (or a private file); first task: “which card for this purchase category + utilisation check”—this is household tooling, not investment advice. Home robots—connect only devices you own; start with status/readbacks before any motion commands; keep a “never unlock / never leave property” rule in the description. Flora—log plants once, then a weekly watering routine that reminds only. Pfp Bot—upload a photo; generate an icon; do not post realistic likenesses (per Marketplace blurb). Keep Personal packs on Personal Bots, not Team seats, when privacy matters [3][6][7].

Product, operations and recruiting

Pack What it does (Marketplace summary)
Proto Bot (Shub Gaur) Ideas → draft PRs with screenshots/video; never merges or publishes without yes.
Stalk Bot (Shub Gaur) Competitor pulse (newsletters, onboarding video, site/pricing/jobs/X); never posts or contacts.
Office Ops Desk (Erika Cabrera) Shipments, facilities, birthdays → digests; never sends without you.

(Recruiting Coordinator is also listed under Featured.)

Implementation ideas: Proto Bot—product brief → draft PR with screenshots/video on a feature branch; first task ends at “ready for review”; never merge. Stalk Bot—seed a competitor list in the description; routine idea: twice-weekly pulse memo (newsletters, pricing, jobs, public X); never contact competitors or post. Office Ops Desk—shipments inbox + facilities calendar; first task: “this week’s digest + draft birthday notes—do not send”; routine: Monday ops digest. Recruiting Coordinator (Featured): ATS + calendar; loop prep only until you approve outreach. Approval across this strip: no merge, no external contact, no send without you [3][6][12].

For the full live list—and any pack added after this fetch—open x.ai/bot/marketplace [3].

Docs use-case roles and Guides (supporting note)

Official use-case pages offer eight starter role recipes—not Marketplace rows and not locked product types: Sales Outbound, Talent Scout, Paid Media, Expense Manager, Product Performance, Bug Reproduction, Account Health, and Chief of Staff. Each starter ends at a review point—no unsupervised send, purchase or production change [12].

SpaceXAI Guides are playbooks, not product enums: six-seat marketing teams [9]; Inbox/Calendar/Intel/Bot Boss work seats [10]; engineering domain fleets plus ops and cloud-agent loops [13]. Index: x.ai/bot/guides. Third-party tutorials sometimes amplify Researcher/Writer/Publisher stacks—useful patterns only, not official types [14].

Caveats / What this does not prove

  • No closed official enum of Research / Writer / Publisher / Social / Game as product types. Those names are jobs, Guide seats or Marketplace packs [2][4][3].
  • Marketplace inventory moves. Counts and Featured rows here are 7 October 2026 only [3].
  • Implementation ideas above are practical suggestions for tools, first tasks and routines—not SpaceXAI product documentation.
  • No billing or weekly usage numbers in this piece—re-check Plans & billing before any limit claims.
  • Template ≠ clone. Add does not transfer computer, logins, secrets or full history [2][8].
  • Oct 7 multi-model routing is not a bot-type list; cite Musk’s post as runtime intent, not taxonomy [11].

The Bottom Line

If you came looking for a fixed menu of “Grok Bot types,” the honest answer is thin: Personal, Team, and Template—then whatever jobs you write or install. The practical path is the how-to spine above: create one clear Personal job, Add Marketplace packs as copies, pick Team vs Template deliberately, staff group chats for visible handoffs, save skills and routines once the process is stable, and keep newsroom publish behind human approval. The install catalog is the Marketplace—all Featured, Engineering, Sales, Marketing, Design, Personal, and Product/Ops/Recruiting packs listed here as of 7 October 2026, with practical ideas for how to run each category [3]. Beyond the store, use the eight docs use-case starters and the Guides playbooks—and re-fetch Marketplace before you republish any pack list.

Sources

  1. SpaceXAI, Grok Bot overview. https://docs.x.ai/grok-bot/overview
  2. SpaceXAI, Create and manage Bots. https://docs.x.ai/grok-bot/bots
  3. SpaceXAI, Grok Bot Marketplace (catalog fetched 7 October 2026). https://x.ai/bot/marketplace
  4. SpaceXAI, Message and collaborate. https://docs.x.ai/grok-bot/chat-and-collaboration
  5. SpaceXAI, Get started with Grok Bot. https://docs.x.ai/grok-bot/get-started
  6. SpaceXAI, Skills, routines and automations. https://docs.x.ai/grok-bot/skills-routines-and-automations
  7. SpaceXAI, Team Bots. https://docs.x.ai/grok-bot/team-bots
  8. SpaceXAI, Templates for Grok Bot (Guide). https://x.ai/bot/guides/templates-for-grok-bot
  9. Josh Kim, Grok Bot for marketing (SpaceXAI Guide), 24 September 2026. https://x.ai/bot/guides/grok-bot-for-marketing
  10. Josh Kim, Grok Bot for work (SpaceXAI Guide). https://x.ai/bot/guides/grok-bot-for-work
  11. Elon Musk, post on X regarding Grok Bot back-end model policy, 7 October 2026. https://x.com/elonmusk/status/2107724314451878104
  12. SpaceXAI, Use cases. https://docs.x.ai/grok-bot/use-cases (mirrored at https://cursor.com/docs/grok-bot/use-cases)
  13. Lingxi Li, Grok Bot for engineering (SpaceXAI Guide). https://x.ai/bot/guides/grok-bot-for-engineering
  14. Postiz, third-party tutorial on Grok Bot content-team patterns (not an official type list). https://postiz.com/blog/grok-bot-ai-agents-content-team-postiz
  15. SpaceXAI, Introducing Grok Bot. https://x.ai/news/introducing-grok-bot
  16. SpaceXAI / Cursor, Grok Bot docs hub. https://cursor.com/docs/grok-bot
  17. SpaceXAI, Guides index. https://x.ai/bot/guides
TSN
TSNhttps://tsnmedia.org/
Welcome to TSN. I'm a data analyst who spent two decades mastering traditional analytics—then went all-in on AI. Here you'll find practical implementation guides, career transition advice, and the news that actually matters for deploying AI in enterprise. No hype. Just what works.

Related articles

Recent articles