The product · Private beta
Lumnya. Your store's digital workforce.
A customer-facing AI that sells and supports shoppers around the clock, and a back office of specialist agents that analyse, forecast and recommend. Everything that matters waits for the owner.
01 · The platform
Four parts. One workforce.
Three layers that work as one team, sharing a single memory of your store.
- Front of house Customer AI Sells and supports shoppers on the storefront, around the clock, grounded in your policies.
- The manager Lumnya Ask in plain language, get answers from live data, approve the actions that matter.
- Back office Agent Workforce Specialists that scan the store on a schedule and propose next steps with their reasoning.
Underneath Multi-model engine Runs on the Claude API: Sonnet 5.5, Opus 5.5 and Fable 5.1, routed by task
The operating loop
Recommend. Approve. Act. Remember.
Agents surface a recommendation with its reasoning. You approve the exact change. Lumnya carries it out in the store, then measures what happened and keeps it in memory for the next call.
02 · Customer AI
A storefront assistant that finishes the job.
The front of house. It sells, supports and closes the loop on orders, day and night, inside the rules you set.
Sells
- Product discovery by need, price or skill level
- Guided checkout that creates the order in Shopify
- Discounts negotiated within your limits, tiers never revealed
- Timely nudges from on-page behaviour, not random pop-ups
Supports
- Order tracking from live order data
- Verified cancellations and end-to-end returns
- Answers grounded in your FAQ and policies
- Hand-off to a person when the shopper is frustrated or the AI is unsure
Speaks
- Replies in the shopper's own language
- Widget and dashboard in English, Greek and German
- Waits for the shopper to finish typing, then answers everything
- Remembers what was said, even in long chats
03 · Lumnya, the manager
Tell it in a sentence. Approve in a click.
Ask in plain language, see the exact change before it happens, and keep every decision on the record.
-
Ask
Sales summaries against the prior period, low stock, churn-risk customers, margin changes. Answers arrive as cards with next steps.
-
Act
Prices down to the variant, full or partial refunds, discount codes, one-off customer emails, returns, FAQ, settings, support tickets and reports.
-
Goals in words
Describe an outcome, approve the plan once, and Lumnya tracks it against real numbers. A goal never grants new permissions.
-
Give it a file
Send a supplier's price list as Excel, CSV, PDF or a photo. SKUs are matched, every change shown side by side, ambiguous prices are never guessed.
-
Store watchdog
Hourly checks for stopped orders, a failing storefront widget and unanswered requests for a human, measured against your normal for that hour.
-
What would you do?
A simulation that changes nothing and shows which actions Lumnya could take on its own.
04 · Agent Workforce
Seven specialists, and more on the way.
Specialists that read your store data, find what matters and bring it to you with the reasoning attached.
Agent Reads Brings you Status
Inventory Forecaster Watches stock across the catalog and proposes priority-ranked reorders, with the reasoning behind every number. ReadsStock levels against your thresholds Brings youRanked reorder recommendations Live
Profit Lens Maps the true margin of every product, from the costs the owner enters. ReadsPrices, discounts and your costs Brings youTrue margins and profit signals Live
Liquidation Planner Finds dead stock and drafts bounded clearance plans for approval. ReadsSlow and dead stock Brings youCapped clearance plans Live
Bundle Discovery Finds products that really sell together, from actual order pairs. ReadsReal purchase pairs Brings youBought-together suggestions Live
Demand Radar Hears demand in storefront conversations that the catalog does not cover. ReadsStorefront conversations Brings youUnmet demand signals Live
Revenue Detective Explains every revenue change by its drivers: volume, basket, refunds and mix. ReadsRevenue over time Brings youQuantified driver breakdowns Live
Self-Learning FAQ Turns questions the assistant could not answer into FAQ drafts, one click to approve. ReadsUnanswered questions Brings youFAQ drafts for approval Live
Next
Pricing Optimizer
Price changes worth making, from margins and elasticity.
Next
Churn-Risk Scorer
Customers most likely to stop buying, flagged early.
Next
Sentiment Pulse
Customer mood from reviews and support tickets.
Next
Anomaly Detector
Alerts on every metric you track.
Next
Executive Summary
A Monday-morning briefing on the whole store.
How the agents work
- 01
Scan and record
Each agent works on a schedule, or on demand, and writes down what it understands.
- 02
Share findings
Facts one agent establishes are reused by the others, tagged with who found them and how confident they are.
- 03
Recommend first
Agents start in recommend-only mode. Autonomy is granted per area of the store, never per agent.
- 04
Measure the outcome
Results are checked against real store data from the day after an action, settling over about a week.
05 · Agent Memory
It remembers your store, and only yours.
Everything Lumnya learns stays inside your store, is never pooled with other merchants and is never used to advise a competitor.
-
Tell it once
Rules you give are kept in your own words and treated as firm limits.
-
See what it learned
A summary by category and count, without exposing customer values.
-
Delete with a trace
Remove a rule or a whole memory category. The deletion itself is logged.
-
Retention you control
Clear retention windows decide how long memory is kept.
06 · Trust & safety
The owner stays in control.
Autonomy needs accountability. These guarantees are enforced on the server, not left to the model.
- 01
Store-isolated data
Each store's data, and the intelligence built on it, stays scoped to that store. Cross-store reads are denied for everyone, including the team.
- 02
Confirm-gated actions
Anything that touches money or a customer shows the exact change first and waits for an explicit yes.
- 03
Full audit log
Every write is journaled with what changed and how it was authorised. Reversible actions can be undone in one click.
- 04
Autonomy on a dial
Independence is set per area of the store, inside caps you choose, and can be paused with one switch.
Autonomy dial · per area
Starts at Confirm
Never unattended
- Refunds
- Customer emails
- Return decisions
- Support tickets
- Live chat sent as you
- Security-sensitive settings
These always wait for a person, at every autonomy level. Customer data is processed to serve your shoppers, not to train foundation models, and every connection is encrypted in transit.
07 · Status
Where Lumnya is today.
Features and availability may change during the beta. The public roadmap and changelog are on lumnya.ai.
- Stage
Private beta
Access by waitlist, with invites sent in waves.
- Platform
Shopify first
Deep Shopify integration today. WooCommerce is on the roadmap ballot.
- Languages
EN · EL · DE
English, Greek and German in the widget and dashboard, with more planned.
- Price
Free in beta
The full product, no card. Beta merchants keep a founding rate at launch.
08 · FAQ
Questions, answered plainly.
What is Lumnya, in one breath?
A customer-facing AI that sells and supports shoppers on your storefront around the clock, plus a back office of specialist agents that work on your store data. Anything consequential waits for your confirmation and is logged.
How does it plug into my store?
Lumnya is a Shopify app. The dashboard runs inside your Shopify admin or as a standalone app, and the storefront assistant is switched on with a theme app embed toggle. No code edits, and nothing is written into your theme files.
Will the AI invent discounts or act on its own?
No. Discounts stay within the limits you set. Refunds, order changes and customer emails need your approval, and every action goes into an audit trail.
Can I let Lumnya do some things without asking me?
Yes, but only deliberately. Independence is set per area of the store, and every area starts at confirm-first. Raised levels run within caps you define. Refunds, customer emails, return decisions, support tickets and security-sensitive settings never run unattended.
If Lumnya changes something, can I put it back?
Usually in one click. Every write is journaled with how it was authorised, reversible actions store their own inverse, and irreversible ones are labelled as such.
Which languages does it speak?
The storefront assistant replies in the shopper's language. The widget and dashboard currently ship in English, Greek and German, with more planned.
Is my customers' data used to train AI models?
No. Data is processed to answer your customers, not to train foundation models. Each store’s data sits in its own isolated space and is never pooled with other merchants.
What does it cost right now?
Nothing during the private beta: the full product, no card required. Launch pricing will be announced before general availability, and beta merchants keep a founding rate.
Private beta · Shopify first · EN · EL · DE
Meet Lumnya.
See the product, or join the waitlist for an invite when the next wave of stores opens.