Grok Bot multi-agent setup

Named seats. Shared gates. A 15-minute paste and a 90-minute first run. A Grok Bot multi-agent desk is a fleet, not a drawer of random bots.

Runtime: Grok Bot only · Product: OF-001 — $29 · Install: /agents.md

What “multi-agent” means on a one-person desk

Agents are seats, not employees:

Chief of Staff — priorities, briefs, handoffs
Market Scout — demand signals, angle cards, watch list
Product — SKU cards, pack outlines, honest listing copy
Store Ops — listing fields, fulfillment notes, delivery QA
Research / Sales-draft — optional; still gated; never send
QA — lane, not a seventh personality

Four core seats, one optional specialist, a QA lane. Not twenty.

Templates optimize for more output. A multi-agent desk optimizes for ownership of the irreversible clicks.

Before you create the second bot

Write the shared rule once. Every seat pins the same default deny: SEND / PUBLISH / PAY. See gates and agent gates.

15-minute install order

Open Grok Bot (custom / project / bot instructions)
Create four bots: COS, SCOUT, PROD, OPS (optional fifth: R&S)
Paste one seat per bot — do not dump every seat into one mega-prompt
Pin the always-on gates block in every seat
Save three routines: morning brief → COS; demand scan → SCOUT; weekly SKU pipeline → COS then PROD
Open a tracker (Notion / Sheets / markdown) for money/SKU + demand watch list

OF-001 paste-ready files: 01–07 and 09 in the zip. Canonical order: /agents.md.

First run: 90 minutes

0–15 COS — morning brief → top 3, one kill, one handoff
15–35 SCOUT — demand scan → three angle cards + watch list (≤5 rows)
35–65 PROD — SKU card + honesty pass (no fake proof)
65–85 OPS — stage listing fields + delivery QA (do not publish)
85–90 You — named yes or explicit stop; then you click in the real tool

Handoffs

To: SCOUT
Ask: Demand scan for [niche]. Three angle cards, no invented market sizes.
Done looks like: markdown table + one recommended angle for PROD.

Failure modes

One mega-bot with seven jobs → split seats
Gates in CoS only → pin everywhere
Twenty templates, no cadence → three routines, one SKU in draft per week
Agents that claim they already sent it → never narrate completed SEND/PUBLISH/PAY
Building on ChatGPT/Claude → this setup is Grok Bot only

CTA

OF-001 zip: CoS, Scout, Product, Ops, optional Research/Sales, routines, gates rulebook, 90-min sample, templates, first-week checklist. $29. Checkout: #whop / YOUR_WHOP_CHECKOUT_URL. Support: mike@ownrshp.io.

# Grok Bot multi-agent setup > Named seats. Shared gates. A 15-minute paste and a 90-minute first run. A Grok Bot multi-agent desk is a fleet, not a drawer of random bots. **Runtime:** Grok Bot only · **Product:** [OF-001 — $29](/products/grok-bot-operator-fleet.md) · **Install:** [/agents.md](/agents.md) ## What “multi-agent” means on a one-person desk Agents are seats, not employees: - **Chief of Staff** — priorities, briefs, handoffs - **Market Scout** — demand signals, angle cards, watch list - **Product** — SKU cards, pack outlines, honest listing copy - **Store Ops** — listing fields, fulfillment notes, delivery QA - **Research / Sales-draft** — optional; still gated; never send - **QA** — lane, not a seventh personality Four core seats, one optional specialist, a QA lane. Not twenty. > Templates optimize for more output. A multi-agent desk optimizes for ownership of the irreversible clicks. ## Before you create the second bot Write the shared rule once. Every seat pins the same default deny: **SEND / PUBLISH / PAY**. See [gates](/guides/send-publish-pay-gates.md) and [agent gates](/guides/agent-send-publish-pay-gates.md). ## 15-minute install order 1. Open Grok Bot (custom / project / bot instructions) 2. Create four bots: COS, SCOUT, PROD, OPS (optional fifth: R&S) 3. Paste one seat per bot — do not dump every seat into one mega-prompt 4. Pin the always-on gates block in every seat 5. Save three routines: morning brief → COS; demand scan → SCOUT; weekly SKU pipeline → COS then PROD 6. Open a tracker (Notion / Sheets / markdown) for money/SKU + demand watch list OF-001 paste-ready files: 01–07 and 09 in the zip. Canonical order: [/agents.md](/agents.md). ## First run: 90 minutes - **0–15** COS — morning brief → top 3, one kill, one handoff - **15–35** SCOUT — demand scan → three angle cards + watch list (≤5 rows) - **35–65** PROD — SKU card + honesty pass (no fake proof) - **65–85** OPS — stage listing fields + delivery QA (do not publish) - **85–90** You — named yes or explicit stop; then you click in the real tool ## Handoffs ``` To: SCOUT Ask: Demand scan for [niche]. Three angle cards, no invented market sizes. Done looks like: markdown table + one recommended angle for PROD. ``` ## Failure modes - One mega-bot with seven jobs → split seats - Gates in CoS only → pin everywhere - Twenty templates, no cadence → three routines, one SKU in draft per week - Agents that claim they already sent it → never narrate completed SEND/PUBLISH/PAY - Building on ChatGPT/Claude → this setup is Grok Bot only ## CTA OF-001 zip: CoS, Scout, Product, Ops, optional Research/Sales, routines, gates rulebook, 90-min sample, templates, first-week checklist. **$29.** Checkout: `#whop` / `YOUR_WHOP_CHECKOUT_URL`. Support: mike@ownrshp.io.