Motir-AI plans work directly INTO external PM tools, creating and updating real items in the vendor's own system while keeping a synced copy in motir-core — the write-back / two-way counterpart of the importer (Issue importer), which is pull-only.
Depends on the importer. Where a vendor is already connected, the write-back reuses the per-vendor OAuth app, API client, and stored connection the importer established. That importer lives in Epic 7: AI Planning Layer (Epic 7), so this epic is blocked_by Epic 7 at the epic level (no cross-epic story edges — the dependency is expressed epic→epic). It also needs, per vendor: write scope requested at authorize time (the same OAuth app is reused — scope is per-authorization, not baked in), and a webhook receiver so the vendor pushes changes back to the motir-core copy.
Writes happen on behalf of the connecting user (their own credential), so items are authored by them — no app-identity / client-credentials tokens needed.
Researched 2026-07-27 (decision ladder rung 1 — checked against each vendor's own product + API docs, not asserted from memory). The finding that shapes this epic: each Chinese tech giant builds its own R&D-management platform and sells it externally, so "support China" is not one integration but four separate vendors — there is no Jira-equivalent single incumbent.
| Giant | Tool it uses and sells | Evidence |
|---|---|---|
| Tencent | TAPD (腾讯敏捷产品研发平台) | tapd.cn — Tencent's own agile R&D platform, sold externally |
| ByteDance | 飞书项目 / Feishu Project (Meego) | marketed as 抖音同款项目管理工具 — the tool Douyin itself runs on |
| Alibaba | 云效 / Yunxiao (Alibaba Cloud DevOps) | Alibaba Cloud's R&D-efficiency platform, built on the group's own R&D practice |
| Huawei | CodeArts (ex-DevCloud); PM module CodeArts Req (ex-ProjectMan) | Huawei Cloud one-stop R&D platform, api-projectman API namespace |
Considered and DEFERRED — explicit disposition, nothing dropped silently: PingCode, ONES, 禅道 ZenTao, CODING (Tencent Cloud), Teambition (Alibaba-owned since 2019), DingTalk. These are real Chinese-market vendors, but they are independent-SaaS / SME / self-hosted plays rather than what the giants run on — the framing this epic was scoped to. ZenTao is the strongest of the deferred set (open-source, self-hosted, a large Chinese SME install base, token API at /api.php/v1/tokens) and is the natural China analogue of the Plane story; add it as a second wave once the four above land.
7.16.14 — OAuth is the SOLE live-source auth recorded OAuth connect as the only live-source auth and removed the paste-a-token path. Of the four Chinese vendors only TAPD has an OAuth 2.0 authorization-code flow; the other three authenticate differently (Yunxiao = a user-created personal access token, CodeArts = Huawei Cloud IAM / AK-SK, Feishu Project = a plugin the customer installs into their space). That ADR therefore has to be amended before three of these four can be connected at all — the 云效 Yunxiao story owns that amendment plus the non-OAuth credential substrate, and the Feishu Project + CodeArts stories are blocked_by it.
Plan in Jira is the foundation: it owns the shared write-back engine (plan → vendor mapping, the external-id mirror map, push/update, conflict + webhook sync), so every other vendor story is blocked_by it and adds only its own adapter. The two China roots are TAPD (OAuth, reuses the shipped substrate as-is) and 云效 Yunxiao (owns the non-OAuth credential substrate); Feishu Project and CodeArts follow Yunxiao.
This epic owns the WRITE direction only. The read / import direction for the Chinese vendors is NOT in scope here — it stays with the importer (Issue importer) in Epic 7; each story below states the hand-off explicitly. Each China story does bring the vendor connection it needs (none of the four is connected today — the importer ships connectors for Jira, Linear, GitHub, Plane and CSV only), built on the shipped connector interface and encrypted identity store.
Lowest priority; intentionally not yet expanded into subtasks — plan each story when it's picked up.