Monadops
Roadmap and architecture
Prepared for Huseyin T. · 27 September 2026
Valid 30 days · Mike Navitsky
Proposal · AI operating system for a new Shopify brand
Launch day: your Shopify store runs on AI. End of January: your whole org chart.
You want your Shopify store run by the AI company in your document. This plan gets you selling first and adds agents fast. Shopify is connected in week 1, and something new goes live every week after that. On launch day the store, Customer Experience, CRM, and Marketing all run on agents. The rest of your chart is live by the end of January, then we keep improving it.
6 weeks
To launch, with a live win every week. Shopify plus the three departments you asked for, on 26 of your 41 tools.
End of Jan
The rest of your chart: Executive Assistant, Growth, CFO, COO, CTO. About eleven weeks after launch.
$18,000
Launch build. Six weekly Upwork milestones of $3,000, each released when you see it working.
$27,000
November to January build, same weekly rate. Then a monthly plan that keeps improving it.
Your Shopify, run by agents
Shopify is item 1 on your list: store, orders, products, customers, inventory. It is also where every other agent gets its facts. The store agent connects in week 1 and runs the store side before your first order. Here is what it does in each of those five areas.
Store Before launch, then every morning
Runs on its own
- A launch checklist in Slack every morning until launch. It covers products, prices, weights, SKUs, images and alt text, SEO fields, shipping rates, taxes, policy pages, the contact page, pixels, and a test order.
- Each missing item becomes a ClickUp task with the fix written out.
- After launch, a 7am Shopify report: orders, revenue, average order, conversion rate, top products, refunds, and discount codes used.
- Search Console from the day the store opens: pages Google has not indexed, and the searches that bring shoppers in.
Waits for your tap
- Any change to store settings, the theme, or navigation.
You stop doing
- Clicking through the admin to find what is broken or missing before launch.
- Pulling yesterday's numbers by hand.
Products
Runs on its own
- Drafts product pages, FAQs, how-to-use steps, SEO titles and descriptions, and image alt text from your product facts and brand book.
- Claims check on every draft. It flags wording that Meta, Google, or TikTok reject, or that you cannot back up, like "removes wrinkles permanently".
- Watches the Merchant Center feed. It explains each disapproval and drafts the fix.
Waits for your tap
- Publishing any product copy.
- Any price or compare-at price change.
- New collections and bundles.
Orders
Runs on its own
- Watches every order: not shipped on time, tracking not moving, or flagged high risk by Shopify.
- Tags orders as first order, repeat, subscription, gift, or VIP.
- Answers "where is my order" in Gorgias with the live status.
Waits for your tap
- Address changes and cancels before shipping. The agent checks the order has not shipped. You tap, it edits the order and replies to the customer.
- Holding or releasing a high-risk order.
- Refunds and replacements. Always.
Customers
Runs on its own
- Tags customers from what they do: first-time, repeat, lapsed, VIP, subscriber, refund history. Klaviyo segments use the same tags.
- One record per customer in the brain with orders, tickets, reviews, and emails. Every agent sees the whole history before it acts.
Waits for your tap
- Any discount code or store credit.
- Editing or merging a customer record.
Inventory
Runs on its own
- Stock and days of cover for every SKU, from sales so far.
- An alert when a SKU drops below the days of cover you set.
Waits for your tap
- A reorder quantity, proposed with the math. You tap and it becomes a ClickUp task.
- Pausing ads or the Shopping feed for a SKU about to sell out.
How it connectsA custom app in your Shopify admin with only the permissions each job needs. Reads go direct. Every write, even a tag, goes through the approval gate and is logged with a way to undo it.
Who ships your ordersA 3PL, your own team, or the supplier. Tell me in week 1. The order watch connects to it in week 4. I run a 3PL integration for my own brand today.
Already done on my own storeMy operations agent reads and writes Shopify orders, fulfillments, tags, notes, and prices through the same kind of gate every day.
A live win every week to launch
Nothing waits for launch day. Each week ends with something working in your store or your Slack, and each week is its own Upwork milestone. If kickoff is Monday 5 October, launch is the week of 9 November.
Week 15 Oct
Jarvis and your store auditServer, brain, approval gate, and Jarvis in your Slack. Shopify connected. Search Console verified on your domain. Decisions locked: Recharge or Skio, who ships, brand voice source, and your product costs.LiveA morning brief in Slack with the launch checklist, your calendar, and overdue ClickUp tasks.
$3,000
Week 212 Oct
Product pages and the Shopping feedProduct pages, FAQs, how-to-use steps, and SEO fields drafted with the claims check. Merchant Center account and feed built and checked. Clarity added to the store.LiveSignup on your coming-soon page, synced to Klaviyo, and the welcome flow, so your list grows before launch.
$3,000
Week 319 Oct
Support ready before the first orderGorgias connected. Help-center answers and macros written from your product facts. AfterShip tracking in every order-status reply.LiveSupport agent answering test tickets end to end, with the approval queue in Slack.
$3,000
Week 426 Oct
Every order and every SKU watchedOrder watch, address change and cancel on your tap, stock alerts. Klaviyo launch flows: abandoned checkout, browse, post-purchase with how-to-use steps, and the Okendo review request. Recharge or Skio if you launch with subscriptions. A first Black Friday stock check from your planned spend, while there is still time to reorder.LiveOrder and stock watch on your test orders. Flows tested and ready.
$3,000
Week 52 Nov
Store opens, ads readyThe store password comes off with ads still off, so Google can review the Shopping feed and read the sitemap before launch. Meta, Google, and TikTok connected. Pixels, Littledata server-side events, the Triple Whale pixel, and the KnoCommerce survey installed and checked with a real test order. Launch campaigns drafted for your tap. Black Friday offer and calendar drafted, so your designer has time for the creatives.LiveThe 7am report with CAC, profit per order, and breakeven ROAS. Spend alerts armed.
$3,000
Week 69 Nov
LaunchFull dry run with test orders, from ad click to delivery email. Command center v1. Runbooks written.LiveYou launch. Shopping ads can run from day one. A daily note on what shoppers do on the store. I watch every agent daily through the first week.
$3,000
Six fixed-price milestones on Upwork, total $18,000. Each is funded when it starts and released when it is live and you have seen it work. If a week slips because of me, it is not released until it is live. If it slips because an account or an answer is late on your side, the date moves and the price does not. Assumes your Shopify store, Gorgias, Klaviyo, ad accounts, and Slack exist by the end of week 1, Triple Whale, Littledata, and KnoCommerce by week 5, and someone on your side answers questions the same day. Google does not review a store behind a password, which is why the store opens in week 5. If you want it closed until launch day, Shopping ads start a few days after launch instead. Your designer or Replo owns the theme and the visuals.
Launch day: the three departments you asked for
The point is fewer things on your desk. Anything reversible or read-only runs on its own from the first day. Approvals stay only where money moves or a customer is upset. Each approval is one tap in Slack, and every tap trains the rule that removes the next one.
Customer Experience
Support agent on Gorgias
Runs on its own
- Reads, tags, and routes every ticket, email and chat.
- Answers "where is my order" when the parcel is on time, with tracking from AfterShip.
- Closes spam and auto-replies.
- Friday digest: top complaints, top questions, fixes to make.
- Negative reviews in Okendo land in Slack with the order history.
Waits for your tap
- Refunds, replacements, discounts, upset customers. Always.
- Every other ticket type until it has 50 approved drafts with no edits. Then it runs on its own.
You stop doing
- Answering order-status and product questions by hand.
Gorgias · Shopify · AfterShip · Okendo · Recharge or Skio · Slack
CRM & Retention
Klaviyo flows plus the retention agent
Runs on its own
- Welcome, abandoned checkout, post-purchase, and winback flows send by themselves once tested.
- Subscription flows (failed payment, cancel save) if you launch with Recharge or Skio.
- Segments rebuilt every night: first-time, repeat, lapsed, VIP, subscriber.
- Morning list of who is due to reorder and who went quiet.
Waits for your tap
- One-off campaign sends. The agent proposes segment, copy, and send time. You tap, it schedules.
- Any discount code.
You stop doing
- Building segments and deciding who to email this week.
Klaviyo · Shopify · Recharge or Skio · Slack
Marketing
Paid media, content, and research agents
Runs on its own
- 7am report across Meta, Google, TikTok, and GA4: spend, revenue, ROAS by campaign.
- Alerts: spend spikes, disapproved ads, Merchant Center feed errors.
- Pauses an ad that spent past your cap with zero sales. Logged, undoable.
- Weekly competitor brief from Meta Ad Library and TikTok Creative Center, filed in ClickUp.
- Ad copy, hooks, and UGC scripts drafted into a queue from your brand book and real customer words.
Waits for your tap
- Any budget or bid change. Any new campaign. Publishing copy.
You stop doing
- Opening four ad dashboards every morning.
Meta Ads · Google Ads · TikTok for Business · Merchant Center · GA4 · Ad Library · Creative Center · ClickUp
Measured from the first visitor
Some data only exists if the tool is on when it happens. A pixel counts only the visits after it is installed. A survey asks only the orders placed after it is live. So the tracking goes in before launch, and Black Friday gets measured.
Where sales come fromLittledata sends server-side purchase events to Meta, Google, and GA4. The Triple Whale pixel and the KnoCommerce "how did you hear about us" survey run from the first order. Survey answers are in the weekly digest from launch week. Full attribution reports come in December, on data collected since launch day.
What shoppers doMicrosoft Clarity records sessions from week 2. From launch, a daily note on rage clicks, dead clicks, and scroll depth by page, next to where shoppers drop off in the Shopify funnel. You watch the recordings in Clarity.
Your money per orderFrom launch day the 7am report shows CAC, profit per order after product cost, shipping, and fees, and the breakeven ROAS for each channel. It uses the costs you give in week 1.
Jarvis sits above all three in Slack. It posts one morning brief grouped by department, routes your requests, and escalates anything an agent is unsure about. A watchdog checks every agent ran and every integration answered, and tells you if not.
How you will know it works
The command center tracks these from day one, per brand: share of tickets closed without a human, first-reply time, approvals per day and how often you edit a draft, Klaviyo flow revenue, ROAS by channel, and model cost per order. If an agent is not earning its place, the numbers show it within weeks.
After launch: the whole chart by end of January
You start selling in November. The rest of your chart goes live in three short builds, one Upwork milestone each, so you can stop after any of them. Black Friday comes first, because it is two weeks after launch. Anything read-only or easy to connect is already in the launch build. What waits needs traffic, sales history, or write access to your store.
Workstream
Oct
Nov
Dec
Jan
Feb
Launch buildShopify + 3 departments
6 weekly wins, launch
NovemberBlack Friday, EA, creative
Black Friday
DecemberAttribution, research, tests
Growth + money
JanuaryInfluencers, COO, web dev
Whole company
From FebruaryOperator plan, Brand 2
Improve, Brand 2
Launch · 5 Oct to 13 Nov
Selling on agents
- Shopify store agent
- Customer Experience, CRM & Retention, Marketing
- Jarvis, brain, gate, command center v1
- 26 tools, with tracking on from the first visitor
$18,000 · six weekly milestones
November · 16 Nov to 1 Dec
Black Friday
- Week 7: the Black Friday plan tuned with real launch data. Emails, ad copy, and stock, all on your tap.
- Week 7: creative performance, which ad and hook sells, in the weekly report.
- 23 Nov to 1 Dec: no changes to live agents. Hourly watch on orders, stock, and ad spend.
- During the freeze, work that does not touch the store: Executive Assistant on Gmail and Drive, Foreplay swipe files, CRO research.
$6,000 · two weeks
December · 2 to 18 Dec
Growth and money
- Attribution reports from Triple Whale, KnoCommerce, and Littledata, on data since launch day
- AI CFO: LTV, BI, financial analyst
- Market research: Semrush, Similarweb, Reddit. Creator discovery in Modash.
- Motion creative reports
- AI Growth Director: Intelligems tests and Replo pages, on approval
$9,000 · three weeks
January · 4 to 29 Jan
Whole company
- Influencers and affiliates: outreach and impact.com, on approval, starting from the December creator list
- AI COO: product research, suppliers, procurement on approval
- Forecasting and cash flow (needs your accounting access)
- Web dev agents on a copy of your theme, every deploy on approval
$12,000 · four weeks
Dates assume kickoff on 5 October 2026 and move with your launch date. No build work from 21 December to 1 January. The agents keep running and I keep watching them. From February, the Operator plan keeps it improving: new tests, new agents, and fewer approvals as each agent earns it. Brand 2 can start in February. You can reorder anything between builds.
What you asked for, what you get
| Your document asks for | When | How |
|---|
| Shopify: store, orders, products, customers, inventory | Launch | Store agent connected in week 1. Every order and SKU watched from week 4. Details in the Shopify section above. |
| Marketing, Support, and CRM on AI agents from day one | Launch | Live by launch week, each with its own list of what it does without you. |
| The full organization: Jarvis, Executive Assistant, seven directors, their managers and specialists | End of Jan | 52 of the 117 roles in your chart are live at launch. The rest go live in November, December, and January. Your chart is below with every role marked. |
| 41 tools and integrations | 26 at launch | 26 at launch, including the infrastructure items. 11 more by the end of January. Four are handled differently, explained below. |
| Governance: autonomous, supervised, approval required | Launch | Your three levels, one rule per action, stored in the brain. Autonomy grows as each action type proves itself in the logs. |
| Company Second Brain with role-based access | Launch | Postgres + pgvector with your twelve knowledge areas. Each agent reads and writes only its own areas. Every record carries a brand tag. |
| Command-center dashboard | 8 of 10 | Eight of your ten dashboard items at launch, including the agent roster and the integrations page, since the watchdog checks every integration from day one. Memory editor in November. Experiments in December. |
| Cross-department learning loop | 6 of 8 steps | Your eight-step example runs at launch except the page change (December) and the packaging review (January). |
| Same system on your other brands | Built for it | Multi-brand from week 1. Each extra brand is $6,000 and two to three weeks, from February. |
| Creative velocity, image and video agents | Partly | Agents write briefs, hooks, scripts, and ad copy fast. The images and video stay with your designer. I do not promise AI-made creatives. |
Your org chart, month by month
This is your chart from your document, every role in it. In the build, each specialist is a skill inside its department agent, not a separate server. You still see every role: each action in the log is tagged with the role that did it. Same reporting lines, far fewer moving parts, lower running cost.
Launch52 roles
November11 · 16 Nov to 1 Dec
December18 · 2 to 18 Dec
January19 · 4 to 29 Jan
Operator11 · me, with the watchdog
Human-led6 · visual design, product testing
Jarvis + Executive AssistantFounder layer5 of 12 at launch
Jarvis
Router + morning briefEscalationsWatchdog
Executive Assistant
EmailCalendarMeeting prepFollow-upDocumentsNotesResearch assistantPersonal tasksReminders
AI CMOMarketing12 of 23 at launch
Paid Acquisition
Meta AdsGoogle AdsTikTok AdsMedia buying analyst
Creative Director
Creative strategy (briefs)Ad copywriterUGC scriptsImage creativeVideo creativeCreative performance
Market Intelligence
Competitor researchAd Library researchTrend researchOffer intelligenceConsumer research
Influencer / Affiliate
Creator discoveryOutreachPartnershipsPerformance tracking
Brand Manager
Content (text)Brand consistencyBrand strategySocial media
AI CX DirectorCustomer Experience12 of 12 at launch
Customer Support
Email supportChat supportOrder supportProduct support
Customer Intelligence
Review miningComplaint analysisCustomer feedbackVoice of customer
Reputation
ReviewsReputation monitoring
Retention / Save
Refund preventionCancellation prevention
AI CRM DirectorRetention9 of 11 at launch
Email Marketing
Email copyCampaignsEmail design
Lifecycle
WelcomeAbandoned cartPost-purchaseWin-backSubscription retention
Retention Intelligence
SegmentationRetention optimizationLTV analyst
AI Growth DirectorWebsite + Funnels2 of 19 at launch
CRO
Funnel analysisHeatmap / behaviorCRO researchA/B testingCheckout optimization
Offer + Funnel
Funnel researchOffer creationUpsellsBundlesPricing research
UX/UI
UX researchMobile optimizationLanding page designUI design
Web Development
Shopify developerFrontendBackend / integrationQA testingDeployment
AI COOOperations + Product5 of 13 at launch
Project Manager
Task coordinationDeadline monitoringFollow-up
Supply Chain
InventoryLogisticsSupplier researchSupplier commsProcurement
Product Development
Product researchCompetitor productsProduct specsProduct testing
Operations Intelligence
Operations intelligence
AI CFOFinance + BI3 of 9 at launch
Ecommerce Economics
ROASCACLTVContribution margin
Business Intelligence
BI agent
Financial Analysis
Financial analystForecastingCash flowExpense monitoring
AI CTOAI Infrastructure4 of 18 at launch
AI Infrastructure
Agent infrastructureServer / VPSDatabase
AI Workforce Engineer
Agent evaluationAgent creationPrompt / contextAgent optimization
Integration Manager
APIMCP / connectorsAutomation
Knowledge Engineer
MemoryRAGData qualityKnowledge graph
AI Skills + Research
New tools + models briefGitHub researchNew agent skillsInternal skills
Company Second BrainShared knowledge layerLive at launch
CompanyFounderProductCustomerDecisionsSOPsBrandCompetitorsHistorical dataAgent learningsResultsExperiments
Access control at launch: each agent can read and write only its own areas. Experiments fill up when A/B testing starts in December.
All 41 tools, by month
What actually runs on your server
Not a chat window and not my laptop. One VPS in your name, Ubuntu and Docker. Each agent is its own container running Hermes Agent, an open-source (MIT) agent runtime from Nous Research, under a process supervisor. It runs 24 hours a day and restarts itself in seconds if it crashes. Each container has its Slack connection, a scheduler for its timed jobs, and its own tool servers (MCP) for the platforms it works with.
You + teamSlack: ask, approve, correct. Command center in the browser.
One VPS · Docker · your account
JarvisRouter, morning brief
Store agentShopify orders, products, customers, stock
Support agentGorgias, Okendo, AfterShip
Retention agentKlaviyo, subscriptions
Paid media agentAds, GA4, Merchant Center
Content + research agentsCopy queue, competitor briefs
Approval gateOne executor for every write. Classifies, asks in Slack if needed, runs, logs, keeps the undo path.
BrainPostgres + pgvector. Rules, facts, action log, brand tag.
Command centerApprovals, activity, failures, KPIs, spend.
Models by APIOpenAI, Anthropic, Gemini. Set per agent, your keys.
Each agent is one Docker container with Hermes Agent and a supervisor. Runs 24/7, restarts itself.
Your platforms
- Shopify
- Gorgias · Okendo
- Klaviyo
- AfterShip
- Recharge or Skio
- Meta · Google · TikTok Ads
- Merchant Center · GA4
- Search Console · Clarity
- Littledata · Triple Whale · KnoCommerce
- Ad Library · Creative Center
- ClickUp · Slack
Reads go direct. Writes go only through the gate.
The pieces, plainly
- Agents: Hermes Agent containers, one per agent, supervised, 24/7. Slack gateway, scheduler, and MCP tool servers are built into the runtime.
- Models: a setting per agent, not a rewrite. GPT-5.5 for routing and reports, Claude for customer-facing drafts to start. Billed to your own API keys.
- Executor: one small service in front of every write. Rules live in the brain, so you change them without a deploy.
- Brain: Postgres + pgvector, nightly backup to S3-compatible storage.
- Command center: a web app on the same box, behind a Cloudflare Access login.
What is not part of the runtime
- My laptop. Claude Code and Codex are my build tools. Nothing that runs depends on them or on me being online.
- Any account I own. VPS, GitHub, API keys, Slack app: all yours. Revoke my access in a minute and it keeps running.
- No-code glue. The agents and tool servers are code in your GitHub, versioned and testable.
This is the setup I run for my own ecommerce brand today: five always-on Hermes agent containers on one VPS, an operations agent with an approval-gated executor on Shopify and the 3PL, an email agent, and a shared Postgres + pgvector brain. Nine scheduled production agents, 62 automation scripts, four sales channels.
A-Z architecture
Top to bottom. You talk to Jarvis. Jarvis routes to a department. Department agents propose actions. The gate decides whether an action runs now or waits for you. Approved actions go to your tools. Everything is written to the brain, which every agent reads before its next task.
A · YouFounder layer
Founder and teamSlack. Ask, approve, correct.
Command centerApprovals · activity · failures · KPIs · spend
November: Executive AssistantGmail · Drive. Your calendar is in the brief from week 1.
B · JarvisChief of staff
JarvisRoutes · morning brief by department · escalations
WatchdogEvery agent ran, every integration answered
C · DirectorsDepartment agents
Store · AI COO
Store agentLaunch checklist · 7am Shopify report
Orders + stockOrder watch · tags · days of cover · reorder proposals
AI CMO
Paid media agentReport · alerts · pause and scale proposals
Content agentAd, email, and page copy. Text only.
Research agentCompetitor ads · weekly brief
AI CX Director
Support agentTickets · order and tracking lookup · replies
Reviews + digestOkendo watch · Friday digest
AI CRM Director
Klaviyo flowsWelcome · abandoned · post-purchase · winback · subscription
Retention agentReorder · winback · segments · one proposal a day
Dec to Jan: AI Growth Director
AI CFO
AI COO
AI CTO skills
D · Approval gateGovernance
1 · ProposeAgent states the action, the reason, and the data
2 · ClassifyAutonomous, supervised, or approval, from the rules
3 · DecideYou tap yes or no in Slack, or it runs if the rule allows
4 · Run + recordOne executor runs it, logs it, keeps the undo path
E · Your tools26 at launch
Shopify
Gorgias
Klaviyo
Okendo
AfterShip
Recharge or Skio
Meta Ads
Google Ads
TikTok for Business
Merchant Center
GA4
Meta Ad Library
TikTok Creative Center
Search Console
Clarity
Littledata
Triple Whale
KnoCommerce
Google Calendar
Slack
ClickUp
Nov to Jan: 11 more, listed above
F · BrainSecond brain
Postgres + pgvectorYour twelve knowledge areas · governance rules · action log · brand tag on every record
Learning loopCorrections saved as rules, read before the next task
BackupsNightly to S3-compatible storage
G · InfrastructureWhere it runs
One VPS, DockerOne container per agent, supervised, 24/7
GitHubAll code, your repository
CloudflareCommand center login and DNS
ModelsOpenAI, Anthropic, Gemini, per agent
Cost meterEvery call, per agent, per brand
Your learning loop, at launch
The example from your document, step by step, with what runs at launch.
Support sees the same product question five times this week and groups it in the Friday digest.Launch
The support agent writes the pattern to the brain as customer intelligence. You confirm it in Slack with one tap.Launch
Jarvis sends it to Marketing and CRM with the ticket examples attached.Launch
Marketing drafts the education copy and a creative brief. Your designer makes the visual.Launch · text
Growth tests a clearer product page. Until December, this is a drafted page-copy change for your approval.December
CRM drafts an education email for new buyers of that product and adds it to the post-purchase flow on your tap.Launch
Operations reviews packaging or instructions. Until January, this is a ClickUp task for you.January
Next month the question count is compared and the result is written back to the brain.Launch
What runs on its own, the rules
Every action an agent can take is classified once. The classification lives in the brain, so you change it without a code change.
| Level | Rule | Examples at launch |
|---|
| Autonomous | Runs, logs, reports. Read-only or reversible. | Reading tickets, orders, and ad accounts. Reports, alerts, and briefs. Tagging, routing, segmenting. On-time order-status replies. Any ticket type that has earned it. |
| Supervised | Runs, shows in the daily digest, can be undone. | Replying to a positive review. Flow emails. Customer notes. Pausing a zero-sale ad inside your cap. |
| Approval | Proposed in Slack with reason and data. Nothing happens until you tap yes. | Refunds, replacements, discounts. Budget or bid changes. Campaign sends. Publishing copy or site changes. Upset customers. Anything the agent is unsure about. |
Price
| Build | Dates | Upwork milestones | Price |
|---|
| Launch: Shopify + three departments | 5 Oct to 13 Nov | Six weekly, $3,000 each | $18,000 |
| November: Black Friday, Executive Assistant | 16 Nov to 1 Dec | One | $6,000 |
| December: growth and money | 2 to 18 Dec | One | $9,000 |
| January: whole company | 4 to 29 Jan | One | $12,000 |
| Whole chart on one brand | Done by 29 Jan | | $45,000 |
Every build is priced at the same weekly rate, $3,000 a week. While a build runs, watching and fixing the live agents is included. Each build is its own fixed-price Upwork milestone, so you can stop after any of them and keep everything built so far.
From February: keep improving
Operator plan · recommended
$3,500 / month
- I watch every agent daily and fix anything that breaks in what I built, at no extra charge.
- New tests, agents, and integrations, about one a week, in the order we set together.
- Approval rules loosened step by step as each agent earns it in the logs.
- Weekly call, weekday Slack support, monthly cost and performance review.
Covers one brand. Each additional brand on the plan adds $1,000 a month. Month to month, cancel any month. Billed as a monthly Upwork contract.
Care plan · lighter option
$1,500 / month
- Monitoring, fixes to what I built, and small tweaks.
- One call a month, weekday Slack support.
- New agents and integrations quoted as fixed-price Upwork milestones, most between $1,500 and $6,000.
Right if the system is enough for a while and someone in-house owns the day to day.
Prefer to spread the cost? The November to January scope can also run on the Operator plan, about one item a week, starting after launch. That finishes around April for about $39,000 all in, including six months of the plan, instead of $45,000 by the end of January.
Your other brands
The platform is multi-brand from week 1: one VPS, one brain with a brand tag on every record, one command center with a brand switcher. Rolling out a brand means connecting its accounts, loading its brand book and product facts, running the agents supervised for a week, then letting the proven ones run. Budget $6,000 per additional brand, two to three weeks each, as its own Upwork milestone. Best start is February, once your first brand has the whole chart. A brand that already runs Klaviyo and Gorgias takes the longer end, because we map what exists first.
Running costs, paid by you to the providers
| Item | Estimate | Notes |
|---|
| VPS (Docker, Postgres) | $60 to $150 / mo | One box carries several brands. Upgrade the box, not the design. |
| Model usage (OpenAI, Anthropic, Gemini) | $200 to $800 / mo | Estimate. Grows with ticket and order volume. Metered per agent in the command center. |
| Cloudflare, S3-compatible storage | $0 to $20 / mo | Command center login, nightly backups. |
| Your tool subscriptions | as today | Shopify, Klaviyo, Gorgias, ad platforms, and the November to January tools stay on your accounts, in your name. |
Honest notes
Already built against
- Shopify Admin API: orders, fulfillments, order edits, tags, notes, prices, product pages, and theme sections, for my own brand.
- Gorgias, Klaviyo, Meta Ads, Google Ads, GA4, Search Console.
- Hermes Agent containers on a VPS with Slack and Telegram gateways, schedulers, and MCP tool servers.
- An approval-gated executor on Shopify and a 3PL, in production for my own brand.
- Email agents with rule stores, and a shared Postgres + pgvector brain read by several agents at once.
New to me, and how that is handled
- At launch: Okendo, AfterShip, Recharge, Skio, TikTok for Business, ClickUp, Merchant Center, Clarity, Triple Whale, Littledata, KnoCommerce. All have documented APIs, and the last four are app installs plus read-only pulls. They sit in weeks 2 to 5, not week 1. If one does not expose what we need, you hear it in week 1 with the substitute.
- Most November to January tools are new to me too. Each gets a one-day API check before it is scheduled, so a dead end costs a day, not a week.
- Eleven weeks for the rest of the chart is tight for one person. It works because the December tools are read-only data pulls and January builds on tools already connected. If a heavy item slips, like influencer outreach or the web dev agents, that item moves to February. The rest of the plan does not.
- Launch lands about two weeks before Black Friday. That is the first real load test, which is why November opens with a Black Friday plan, a watch, and a change freeze.
One person, start to finish. You get me on the weekly call and in Slack, not an account manager. You own the code, the server, and the data. Cancel the plan and everything keeps running. Anyone who can restart a container and read a log can take it over.
Next step
Reply in this Upwork thread
Three answers and a 45-minute call
Tell me your launch date, who ships your orders, and whether it is Recharge, Skio, or no subscriptions at launch. I send the Week 1 milestone on Upwork and we book the kickoff call through the Upwork room. Bring the brand name and the status of your Shopify store, Klaviyo, and ad accounts. I bring the week 1 blueprint draft and your chart marked month by month, ready to reorder.
If after that call this is not the right build for you, we stop there and you keep the blueprint notes.