SEA Marketplace Publishing APIs  ·  2026-07-07  ·  Draft

Publishing our AI Product Detail Pages into every major SEA marketplace — through official APIs

Lazada, Shopee, TikTok Shop and Tokopedia all expose an API that lets one approved app publish listings into many sellers' shops on their behalf — the exact model our product is built around. The two hardest pieces already live in our codebase.

4 / 4
Marketplaces publishable via official API
~2 months
To first two live (Lazada + Shopee) · ~4–5 months for all four

01 The direct answer — what each API does

MarketplaceOfficial publish APICreate a listingSeller connectsPublish?Phase
Lazada Open Platform (LazOP 2.0) POST /product/create XML payload + image migrate OAuth — seller self-authorizes via link Yes Phase 1
Shopee Open Platform v2 add_itemadd_model + media upload Per-shop OAuth (auth_partner) Yes Phase 1
TikTok Shop Partner Center Open API 202309 POST /product/202309/products Partner app OAuth + shop_cipher Yes Phase 2
Tokopedia Same API as TikTok Shop standalone API retired ~30 Sep 2025 Identical to TikTok Shop Via TikTok Shop Partner Center Yes Phase 2
The strategic finding on Tokopedia
After ByteDance's acquisition of a stake in GoTo/Tokopedia, Tokopedia's standalone API was terminated (~30 Sep 2025) and Indonesian publishing now runs on the same TikTok Shop API. Building TikTok Shop delivers most of Tokopedia for free — the fourth marketplace is a ~1–2 week increment on the third, not a rebuild.

02 Why we can move fast

The hard parts are already built

Publishing is well-trodden plumbing, and we already own the parts that are genuinely hard.

HMAC request signing
The cryptographic signing every one of these APIs requires — implemented for Shopee and Lazada today.
Category & brand sync
We already pull each platform's product taxonomy — ~60% of the read scaffolding a publish flow needs.
Multi-tenant base
Our Organizations → Projects → Products hierarchy and S3 image storage are exactly what a per-seller publishing service sits on.
Honest current state
Today customers publish by downloading an Excel template we generate and uploading it by hand in their Seller Center. It works, but it's manual. API publishing turns that into a one-click "Connect your shop → Publish" flow.

03 What we need to build

Four reusable pieces, then a thin write-path per marketplace

01 "Connect your shop" button
A standard OAuth flow so each seller authorizes us once — no more pasting credentials.
02 Encrypted token vault
A secure multi-tenant store for every seller's connection. A durable platform asset — "connect once, publish forever" — that underpins every future marketplace.
03 Image upload step
None of the four accept our hosted image URLs directly; each image is uploaded to the platform first. A known, one-time build.
04 Create-listing + status
The actual publish, plus recording each listing's ID and moderation outcome so we can update and retry.

None of this is novel R&D. It is defined, estimable engineering.

04 Cost structure

CostWho bears itNotes
Marketplace API accessNone foundNo per-call or usage fee identified on any of the four to confirm
Engineering buildUs~15–20 engineering weeks total for all four
Platform approval / onboardingUs — timeBusiness verification + app review; runs in parallel with engineering
Marketplace commission (Shopee ~8%+, Lazada ~4–9%, Tokopedia ~2–12%)The sellerMarket context only — borne by the customer's shop, not an API cost to us
The point for investors
The marginal cost of publishing a listing via API is effectively zero — the investment is a one-time engineering build, after which each publish is a free API call.

05 The plan & timeline

Phased, value-first — approvals run in parallel

Phase 1 · SEA-native

Lazada + Shopee

Engineering~8–11 wks
Calendar (w/ parallel approvals)~2–2.5 mo
Why firstSellers self-authorize; signing + sync already built
Phase 2 · greenfield

TikTok Shop + Tokopedia

Engineering~7–9 wks
Calendar (w/ parallel approvals)~2–2.5 mo
NoteOne shared API — Tokopedia is a ~1–2 wk increment
Two tracks, in parallel. Engineering and platform approval run at the same time, so calendar time per phase is roughly max(build, approval) — not the sum. Total: ~15–20 engineering weeks (~4–5 months) for all four with a small 2–3 engineer team. Scope is listing publish only; a 1-week de-risking spike is recommended before external dates are committed.

06 Risks & dependencies — the honest column

Confirmed gate
TikTok's 25-seller review gate. A custom app needs mandatory TikTok review once it is Connector-category or reaches 25+ seller authorizations. Mitigation: submit before onboarding the 25th seller.
Lead-time
Approval we don't control. Shopee ISV go-live, Lazada permission, TikTok Partner review — days-to-weeks. Mitigation: start all applications on day 1, in parallel with the build.
Pipeline
Images must be pre-uploaded. No platform takes our S3 URLs (Lazada caps at ≤1 MB). Adds one shared, one-time build.
Moderation
Publish ≠ live. All platforms moderate after create; brand/category gating can reject a listing post-submit. Mitigation: status tracking + reconciliation.
To confirm
A handful of facts. Lazada/TikTok token lifetimes, review durations, and Tokopedia's catalog model — de-risked in a 1-week spike before dates are locked.

07 Talking points

  • The capability is real and supported. Every target marketplace has an official API to publish into many sellers' shops on their behalf — not a workaround, not a scrape.
  • We own the hard core already. HMAC signing and category sync for Shopee and Lazada are done; the rest is well-trodden plumbing.
  • Value ships in phases. One-click API publishing to Lazada + Shopee in ~2–2.5 months, rather than waiting for all four.
  • Smart sequencing compounds. TikTok Shop and Tokopedia are now one API — the fourth marketplace is a ~1–2 week increment on the third.
  • The token vault is a durable asset. "Connect your shop" becomes a one-time seller action underpinning every future marketplace and automated republish.
  • We're disciplined about dependencies. The real long-pole is platform approval — we start it in parallel so it doesn't extend the timeline.