Privacy policy

Note: this English version is a machine translation for reference. The authoritative version is the Chinese one at /privacy/. Last updated: 2026-10-06

A tool that does your bidding shouldn't ask for blind trust. This document explains what data Nomu processes, where it goes, and what we will never see.

One-line summary

Store settings and batch drafts stay local by default · No usage-behaviour tracking; anonymous diagnostic events only when the extension crashes · Capture supports 1688 / Taobao / Tmall / JD.com · There are exactly five AI call types, handed to third-party model providers, and the upstream content of each is spelled out · Product drafts parsed via right-click are stored server-side, but are available only to other devices under your own account.

The Nomu account

Every page of the Nomu extension, the capture drawer, and the floating button all require you to sign in to a Nomu account. Sign-up and sign-in work like this:

• Sign up happens on the website (nomu.kanocifer.chat/register) and needs a username, email, email verification code, and password. Sign in happens on the extension's account page and supports either an email verification code or an email link (click the link in the email to confirm). Neither needs a password.
• The password is submitted over HTTPS to the account service (api.kanocifer.chat) only by the sign-up and password-reset forms on the website. The extension itself neither collects nor stores your password.
• After sign-in, the access and refresh tokens are stored in your browser's local storage and are used only to identify you to Nomu services. Logging out clears them locally.
• The account follows a multi-device model: one account can stay signed in on up to 3 devices, and signing in on a new device does not invalidate existing sessions. Signing out on a device clears only that device's session.
• The Nomu account is fully independent of your Noon sign-in: signing into Nomu does not read or affect your Noon session.
• When you are not signed in, the extension's pages show a sign-in prompt, and the capture drawer's form and submit buttons likewise require you to sign in first. The host product pages themselves (1688 / Taobao / Tmall / JD / noon.com) are not affected — you can browse and view products normally before signing in; only the capture and action features Nomu adds to those pages need an account.

Where data goes

Stays local (in your browser)

The following data is only stored in your browser's local storage and is never uploaded to any server:

• Your store settings (country, PartnerCode, warehouse, quantity, warranty, brand)
• Your store records and the currently active store
• Your product batch drafts
• Nomu account sign-in tokens
• NomuDesign drafts (prompts, reference images, generation history and canvas edit state)

The white-background image processing for product images also runs locally in your browser and does not go through any server.

Sent only to Noon (the content you submit)

When you actively submit a listing, the following data is sent to Noon via your already-signed-in Noon session:

• Product information (title, description, attributes, price, stock)
• Product images (already locally processed to white-background JPEG: width ≥ 660px, aspect ratio ≥ 0.5, ≤ 10MB each)
• Warranty and activation requests

These requests are identical to those you would generate by hand in the Noon seller backend — Nomu just executes them for you.

Sent to the account service (on sign-in / sign-up)

On sign-in and sign-up, the following information is submitted over HTTPS to the Nomu account service (api.kanocifer.chat):

• Username and password (on sign-up / password reset, submitted by the nomu.kanocifer.chat website)
• Email and email verification code (on sign-up, code sign-in, email-link sign-in, and password reset)

Nomu AI services (sign-in required)

The following five call types are received by Nomu's cloud service and then passed to third-party model providers (see "Third-party model providers" below for the mapping). Requests carry your account access token for authentication and credit deduction; the token goes only to Nomu's cloud and is never forwarded to a model provider. The upstream content of each is listed below — please pay particular attention to types 4 and 5:

• Translation — uploads the product text to translate (title, description); receives the translation back.
• Prompt optimization — uploads the prompt text; receives the rewritten prompt back.
• Image generation — uploads the prompt, optional reference image base64, and generation parameters (model, size tier); receives the media address of the generated image back (api.kanocifer.chat/v3/media/design/*). Reference images are only used for this generation; they are not retained on the account service side.
• Product parsing (v2/nomu/product-parse) — triggered by the page right-click menu item "AI parse product draft". This is by far the heaviest upstream payload. It uploads, from the page you right-clicked: the page URL, title, description, price and currency, JSON-LD structured data (up to 5 entries), primary images (up to 20), page images (up to 40), the full body text of the page, and a full screenshot of your current viewport (raw base64 JPEG). Parsing is asynchronous: the response returns a job id first.
• Assistant Q&A (v2/knowledge/ask) — see the "Nomu assistant" section below.

For the first three types, beyond what authentication and billing require, requests do not carry your store configuration, batch drafts, or other local data. Product parsing and assistant Q&A necessarily carry content you actively provide — respectively the body text and viewport screenshot of the page you right-clicked, and your questions.

Third-party model providers

Nomu's cloud does not train its own models. The five AI call types are passed to the following third-party providers by type. Your account token, store configuration, batch drafts and sign-in details are never sent to them:

Call type	Provider	Sent to the provider	Returned to you
Translation	DeepSeek	The product title and description to translate	The translation
Prompt optimisation	DeepSeek	The prompt text	The rewritten prompt
Assistant Q&A	DeepSeek	Your question, plus the conversation context of that session	The streamed answer
Product parsing	DeepSeek	Page URL, title, description, price and currency, JSON-LD (up to 5 entries), primary images (up to 20), page images (up to 40), the full body text of the page, a full screenshot of your current viewport (base64 JPEG)	The parsed product draft (also written to your own cloud pool)
Image gen Doubao-*	Volcengine Ark (Doubao Seedream 5.0 / 5.0-pro)	The image prompt, optional reference images (URL or base64), and the size tier	The generated image, stored on Nomu's own media storage
Image gen gpt-image-*	apiyi	The image prompt and optional reference images (multipart upload)	The generated image, stored on Nomu's own media storage

A few notes on image generation:

• Routing follows the model: whichever model you pick in NomuDesign decides who runs it — Doubao-* goes to Ark, gpt-image-* goes to apiyi. The two hold independent keys and can both be in use.
• Reference images are only used for that one generation. We do not retain them on our own servers, but they are sent with the request to the image provider listed above, which performs the generation.
• Results are re-hosted immediately: the image URL a provider returns is temporary (apiyi's expires in about 24 hours), so the cloud copies it straight to Nomu's own media storage (api.kanocifer.chat/v3/media/design/*) and gives the extension only that first-party address.

Upstream retention and training: we do not know how the third-party providers retain logs internally or whether they use requests for training, and we cannot commit on their behalf. Their terms may change at any time. Assess the risk yourself before using the entry points that carry sensitive content (page body text, viewport screenshots, assistant questions, reference images); each provider's terms are whatever its own published privacy policy and terms of service say.

AI credits & consumption history (persisted on the account side)

The account service deducts credits for the five calls above and persists them server-side:

• Credit balance (in cents, 100 cents = 1 unit, credits do not expire)
• Per-transaction consumption history (source / time / cost). Source enum: translate (general translation), nomu_prompt_optimize (prompt optimization), design_generate (generation), plus one independent source each for product parsing and assistant Q&A (exact enum values are whatever the account service returns)

Credits are visible on the account page and can be manually refreshed. To delete your Nomu account and the corresponding credits, contact the author by email (see end of document).

FX service

When listing, the extension calls the FX service to convert local-currency prices. The request only carries the FX query parameters, not your product or store data.

Cloud config sync (sign-in required, manually triggered)

The account page exposes two manual buttons: Upload to cloud and Download to local. When triggered, your store configuration (PartnerCode, store code, memo, warranty, PSKU prefix, etc.) syncs via the account service (api.kanocifer.chat). Only triggered manually by you on the account page — it never runs automatically. The last-synced time is stored locally in your browser and shown on the account page.

Cloud pool (sign-in required)

When the right-click "AI parse product draft" action succeeds, the resulting product draft is saved server-side via the account service (api.kanocifer.chat) so that other devices under your own account can pick it up. The behavior:

• The transport is an account-authenticated sync connection (v3/nomu/sync/ws, which carries your account token at connect time). Pool contents are therefore visible only to you — no other Nomu account can access them.
• On startup the extension pulls the full contents of the pool; when new items appear, the server broadcasts the delta to the online devices under your account.
• Each snapshot carries the writer's device identifier device_id and device name, so you can tell which device wrote it.
• When you "claim" a snapshot on one device, that item is removed from the account's other devices (exclusive claim).

Note that the product draft leaves your machine and is stored on the publisher's server to support your own multi-device use. Do not use right-click parsing on pages containing sensitive information. You can claim or delete pool items yourself in the extension; to purge the full history, contact the author by email (see end of document).

Nomu assistant (sign-in required)

The Nomu assistant is a Q&A assistant backed by a server-side knowledge base. Its upstream and storage behavior:

• Asking: v2/knowledge/ask. Uploads your question text and account access token; the answer is streamed back over SSE.
• Conversation text is stored on the server: locally we keep only the session id, title and last-used time (an "address book"); the conversation body is kept server-side, and the extension reads it back by session_id to display it.
• Knowledge base status: v2/knowledge/status reads ingestion progress; v2/knowledge/ingest triggers an incremental ingestion run. The knowledge base is maintained by the author and is not built from your data.
• The assistant consumes credits; no request is made when signed out or out of credits.

Note that your questions leave your machine, are sent to DeepSeek to run the inference, and are persisted server-side. Do not submit sensitive information to the assistant. To delete a session, contact the author by email (see end of document).

Anonymous diagnostic events

The extension reports one diagnostic event when it throws an uncaught exception, plus one version heartbeat the first time each version runs. These are sent via the account service (api.kanocifer.chat) and need no sign-in. No normal usage is recorded. See the "Anonymous diagnostic events" section below.

What we will never see

• Your Noon account password
• Any Noon-facing keys or OAuth grants
• Any usage-behaviour tracking: feature clicks, page visits, or usage analytics
• Your browsing history or unrelated site cookies
• NomuDesign reference images (we do not retain them on our own servers; they are sent only with the image-generation request to whichever image provider is being used)

Exceptions you should know about

The items below do leave your device and are stored on the server. They are collected here so nothing is missed:

Data	Stored in	Who can see it
Body text and full-viewport screenshot of the page you right-clicked	Nomu AI service (product parsing)	The parsing process
Product draft produced by right-click parsing, plus the writing device's device_id and device name	Cloud pool	Only other devices under your own account
Your assistant questions and answers	Nomu assistant service	Persisted server-side, read back by session_id
The upstream content of the five AI calls above	Third-party model providers	Only to run that inference — see "Third-party model providers"
Email, email verification code, account access token	Account service	Required for account authentication and billing; not forwarded to model providers
Anonymous diagnostic events when the extension crashes	Nomu diagnostics service	Only to diagnose and fix defects

None of the above is shared with third parties: the server side is the publisher's own account service, and access is isolated by account token. The one exception is the "Third-party model providers" row: to complete an AI call you started, its input is submitted to the matching model provider — see "Third-party model providers" above.

Anonymous diagnostic events

To diagnose and fix defects, the extension reports a diagnostic event when it throws an uncaught exception. It also sends one version heartbeat the first time each version runs, solely to count installs per version. No normal usage is recorded — there are no feature-click events and no page-view analytics.

The uploaded fields are strictly limited to: the error summary, the error stack, the extension version, the runtime environment (background / extension page / content script), the browser user agent, and an anonymous install identifier. The server additionally records the request IP address. The extension does not deliberately upload product data, page content, or account information; but the error summary and stack come from the exception itself, so a URL or text fragment involved in the error may travel with it.

The install identifier is a UUID generated randomly on first run. It only links events from the same installation: it is not tied to your Nomu account, not merged across devices, and cannot be used to identify you.

This data is not shared with third parties and is used only to fix defects.

Permissions

Nomu requests the following Chrome permissions:

Permission	Purpose
Storage	Save your store records, settings, batch drafts, and sign-in tokens in the browser
Alarms	Process your listing task queue in the background on schedule
Notifications	Pop a merged reminder when a parse job returns or a shared-pool task is queued (multiple events collapse into one); clicking it opens the corresponding panel
Context menus	Provide "AI parse product draft" in the page right-click menu, for capturing products from your own storefront / brand website. Clicking the menu item counts as an active authorization
activeTab (on-demand)	Only when you actively trigger "capture current page" or click the right-click menu item; temporarily accesses the active tab without pre-authorizing any site
scripting (script injection)	Pair with activeTab; inject the capture script into the active tab when you trigger capture, reading product info from the page

We do not request sensitive permissions like cookies or tabs: the extension does not read browser cookies. Other sites are not accessed by the extension, except when you actively trigger "capture current page" or click the right-click menu item "AI parse product draft" — under activeTab, the active tab is only temporarily accessed (browser internal pages and the Chrome Web Store are always excluded), and authorization expires when you leave the tab.

The right-click menu item appears on any http / https page, so that it covers your own storefront and brand website. Please note its consequences: clicking it submits the current page's body text and a full-viewport screenshot to the Nomu AI service for parsing, and the resulting draft is stored server-side for pickup by other devices under your own account (see above).

Sites accessed

Site	Reason
Noon store (noon.com)	Inject capture tool into the page; read product page info for duplicate / list
Noon partner backend (noon.partners)	Catalog management pages and listing requests
Noon image service	Image and file uploads
Noon hosted image storage (f.nooncdn.com)	Read source product images when duplicating
1688 website and image service (1688.com / alicdn.com)	Capture product info from the page; images fetched via the alicdn CDN
Taobao / Tmall website (item.taobao.com / detail.tmall.com)	Capture product info from the page; images fetched via the alicdn CDN
JD.com website and image service (jd.com / 360buyimg.com)	Product page capture and image fetch
Nomu account service	Account sign-in authentication, AI credit balance and history lookup
Nomu translation service	Text translation
Nomu image generation service	Prompt optimization and product image generation
Nomu media storage	Storage for NomuDesign-generated images
Nomu product parsing service	Asynchronous parsing for the right-click "AI parse product draft" action, and writing the result to the cloud pool
Nomu assistant service	Knowledge-base Q&A, session read-back and ingestion status
DeepSeek	Model inference for translation, prompt optimisation, product parsing and assistant Q&A
Volcengine Ark	Product image generation for Doubao-* models
apiyi	Product image generation for gpt-image-* models
FX service	Convert local-currency prices during listing

The extension's cross-origin access is limited to: noon.partners, alicdn.com, 360buyimg.com, f.nooncdn.com, api.kanocifer.chat. The Noon store, 1688, Taobao / Tmall, and JD.com are not in the list — their product data is read in-page by content scripts in the same origin, scoped to the current product page. Other sites are only temporarily accessed when you actively trigger "capture current page" (see Permissions).

Access to Noon APIs is gated by a URL allow-list. Anything outside is rejected.

The DeepSeek, Volcengine Ark and apiyi rows above are calls made by the cloud service to its model providers, not requests from the extension, so they do not appear in the extension's cross-origin access list.

Changelog

• 2026-10-06 — Added a disclosure of the anonymous diagnostic event data flow. A new "Anonymous diagnostic events" section sets out when an event is sent (one when the extension itself throws an uncaught exception, plus one version heartbeat the first time each version runs), what is uploaded (the error summary, the error stack, the extension version, the runtime environment, the browser user agent and an anonymous install identifier, with the server also recording the request IP), what the install identifier is, and two facts worth stating plainly: there is no switch to turn reporting off, and there is no retention period or automatic cleanup. "Exceptions you should know about" gains a row for the same flow. Previously over-broad absolute wording is narrowed at the same time: "Any tracking, analytics, or telemetry data" under "What we will never see" now reads "Any usage-behaviour tracking: feature clicks, page visits, or usage analytics", so the absence of anonymous diagnostic events is no longer implied.

• 2026-10-03 — Added a "Third-party model providers" section. All five AI call types are handed by the cloud to third-party providers, listed by type with what each receives: translation, prompt optimisation, assistant Q&A and product parsing go to DeepSeek, while image generation is routed by model to Volcengine Ark (Doubao-*) or apiyi (gpt-image-*). Three previously misleading statements are corrected: the "Nomu AI services" section now says the calls are passed on and that the account token is not forwarded; the reference-image entry under "What we will never see" now reads "not retained on our own servers, but sent with the image request to the image provider"; and "Exceptions you need to know about" gains a row for the upstream content of AI calls, which is now the only outbound disclosure. The sites-accessed table gains the three providers, noted as cloud-side calls that are not part of the extension's cross-origin permissions.

• 2026-10-02 — Sign-in gating is now at every entry point: when signed out, the extension's pages, capture drawer and floating button all require an account.

  1. Pages: the ten standalone extension pages — dashboard, tasks, options, nomu-catalog, nomu-design, nomu-agent, nomu-fx, nomu-pricing, duplicate, barcode — are replaced in full by a sign-in wall. The popup keeps its brand header and dock navigation, with the store carousel position replaced by a sign-in card. The capture drawer keeps its shell and close button, with the form and action buttons replaced by a sign-in card. The noon-catalog floating button collapses to a single "Sign in" action.
  2. Account model corrected: the "single session" description is corrected to "multi-device" — one account can stay signed in on up to 5 devices, signing in on a new device does not affect existing ones, and signing out on a device clears only that device.
  3. Host pages are unaffected: 1688 / Taobao / Tmall / JD / noon.com product pages browse normally as before — this is unchanged.

• 2026-09-30 — Three previously undisclosed data flows documented, plus two corrections.

  1. New "Cloud pool" section: the output of the right-click "AI parse product draft" action is saved server-side via the account service and synced across devices under your own account over an account-authenticated connection (v3/nomu/sync/ws). Items carry the writer's device_id and device name, and are removed from the account's other devices once claimed. Not shared with other Nomu accounts.
  2. New "Nomu assistant" section: assistant questions and answers are stored server-side (locally only the session id, title and last-used time), read back by session_id.
  3. AI call types expanded from three to five: added "product parsing" (uploads the page body text and a full-viewport screenshot) and "assistant Q&A"; the credit history source enum gains two corresponding entries.
  4. The permission table drops "Side panel" — the extension does not request that permission — and gains "Context menus", which was missing. Since the right-click menu appears on any http/https page, its consequences are now called out in the permissions section.
  5. The local image-processing spec was corrected from "660×900" to "width ≥ 660px, aspect ratio ≥ 0.5, ≤ 10MB".

• 2026-09-28 — Account method notes updated. Sign-up and password reset now happen on the website (nomu.kanocifer.chat/register, nomu.kanocifer.chat/forgot-password); the extension no longer ships a sign-up form or password sign-in. Sign-in is now the account page's "email verification code" and "email link" pair. The password is only submitted by the website forms; the extension neither collects nor stores it. No other data handling changed.

• 2026-09-13 — Permission tightening. Removed the tabs permission; removed direct-access permissions for the Noon store (noon.com), 1688, Taobao / Tmall, and JD.com. Their product data is now read in-page by content scripts in the same origin; capture scope is unchanged. Added activeTab (on-demand) and scripting (script injection) for "capture current page", effective only when you actively trigger it. Permission table and sites-accessed table updated accordingly.

• 2026-09-11 — Product capture adds Taobao / Tmall and JD.com as new sources. Read behavior is identical to 1688 — only page product info. The sites-accessed table adds the Taobao / Tmall and JD.com rows.

• 2026-09-09 — Cloud config sync added. The account page gains "Upload to cloud" and "Download to local" manual buttons; store configuration syncs via the account service. Only manually triggered; never runs automatically. Last-synced time is stored locally in the browser.

• 2026-09-07 — NomuDesign (product image generation / prompt optimization) and the AI credits system integrated. Addendum: prompts and (optional) reference images are uploaded with generation requests; reference images are not retained server-side. The account service persists credit balance and consumption history. The translation section is upgraded to the AI services section, including image generation and prompt optimization. Permission table gains Alarms. Sites-accessed table gains Nomu image generation service, Nomu media storage, FX service.

• 2026-09-05 — Nomu account system integrated — using translation requires signing in to a Nomu account (supports password and email magic-link sign-in). Added notes for account data (sign-up info, local session tokens, single-session model). Clarified that image white-background processing runs locally in the browser, correcting the previous "image processing is server-side" wording.

• 2026-08-30 — Tightened site-access scope — removed 3 wildcard-overlapping duplicate domains. The Noon image service was narrowed to the actually-used image domains. Image fetching adds a domain allow-list; anything outside is rejected. Alarms permission and Noon hosted image storage were added to the listings.

• 2026-08-28 — Removed the Chrome cookies permission. "Detect current store" was switched to Noon Catalog's own store list API (/_vs/mp/mp-noon-merchant-api/noon-store/list); the extension no longer actively reads browser cookies. Other behavior is unchanged.

• Updates to this policy will be recorded on this page.

Contact

For questions about this policy or data handling, contact: dethe3255@gmail.com.