diff --git a/README.md b/README.md
index 40ad90ff..d771585a 100644
--- a/README.md
+++ b/README.md
@@ -170,27 +170,31 @@ For feature architecture and implementation guidance:
 
 ## Comparison
 
-| Feature | Agent Teams | Vibe Kanban | Aperant | Cursor | Claude Code CLI |
+| Feature | Agent Teams | Gastown | Paperclip | Cursor | Claude Code CLI |
 |---|---|---|---|---|---|
-| **Cross-team communication** | ✅ | ❌ | ❌ | — | ❌ |
-| **Agent-to-agent messaging** | ✅ Native real-time mailbox | ❌ Agents are independent | ❌ Fixed pipeline | ❌ | ✅⚠️ Built-in (no UI) |
-| **Linked tasks** | ✅ Cross-references in messages | ⚠️ Subtasks only | ❌ | ❌ | ❌ |
-| **Session analysis** | ✅ 6-category token tracking | ❌ | ⚠️ Execution logs | ❌ | ❌ |
-| **Task attachments** | ✅ Auto-attach, agents read & attach files | ❌ | ✅ Images + files | ⚠️ Chat session only | ❌ |
-| **Hunk-level review** | ✅ Accept / reject individual hunks | ❌ | ❌ | ✅ | ❌ |
-| **Built-in code editor** | ✅ With Git support | ❌ | ❌ | ✅ Full IDE | ❌ |
-| **Full autonomy** | ✅ Agents create, assign, review tasks end-to-end | ❌ Human manages tasks | ❌ Fixed pipeline | ⚠️ Isolated tasks only | ✅⚠️ (no UI) |
-| **Task dependencies (blocked by)** | ✅ Guaranteed ordering | ❌ | ⚠️ Within plan only | ❌ | ✅⚠️ (no UI, no notifications) |
-| **Review workflow** | ✅ Agents review each other | ❌ | ⚠️ Auto QA pipeline | ❌ | ✅⚠️ (no UI) |
-| **Zero setup** | ✅ | ❌ Config required | ❌ Config required | ✅ | ⚠️ CLI install required |
-| **Kanban board** | ✅ 5 columns, real-time | ✅ | ✅ 6 columns (pipeline) | ❌ | ❌ |
-| **Execution log viewer** | ✅ Tool calls, reasoning, timeline | ❌ | ✅ Phase-based logs | ✅ | ❌ |
-| **Live processes** | ✅ View, stop, open URLs in browser | ❌ | ❌ | ✅ | ❌ |
-| **Per-task code review** | ✅ Accept / reject / comment | ⚠️ PR-level only | ⚠️ File-level only | ✅ BugBot on PRs | ❌ |
-| **Flexible autonomy** | ✅ Granular settings, per-action approval, notifications | ❌ | ⚠️ Plan approval only | ✅ | ✅ |
-| **Git worktree isolation** | ✅ Optional | ⚠️ Mandatory | ⚠️ Mandatory | ✅ | ✅ |
-| **Multi-agent backend** | ✅ Codex, Claude, and 75+ providers | ✅ 6+ agents | ✅ 11 providers | ✅ Multi-model | — |
-| **Price** | **Free** | Free / $30 user/mo | Free | $0–$200/mo | Provider subscription |
+| **Cross-team communication** | ✅ Native cross-team messages | ⚠️ Cross-rig coordination | ⚠️ Company-scoped org work | N/A | ❌ |
+| **Agent-to-agent messaging** | ✅ Native real-time mailbox | ✅ Mailboxes + handoffs | ⚠️ Comments + @mentions | ❌ | ✅ Team mailbox, no UI |
+| **Linked tasks** | ✅ Cross-refs + dependencies | ⚠️ Beads deps + convoys | ✅ Goals, parents, blockers | ❌ | ✅ Shared task list |
+| **Session analysis** | ✅ Task logs + token tracking | ⚠️ Session recall, feed, OTEL | ⚠️ Run transcripts + cost audit | ❌ | ⚠️ Usage command, no UI |
+| **Task attachments** | ✅ Auto-attach, agents read & attach files | ❌ Not task-level | ✅ Docs, attachments, work products | ⚠️ Chat session only | ⚠️ Chat images only |
+| **Hunk-level review** | ✅ Accept / reject individual hunks | ❌ | ❌ Bring your own review | ✅ | ❌ |
+| **Built-in code editor** | ✅ With Git support | ❌ | ❌ Control plane, not editor | ✅ Full IDE | ❌ |
+| **Full autonomy** | ✅ Agents create, assign, review tasks end-to-end | ✅ Mayor, convoys, recovery | ✅ Heartbeats + governance | ⚠️ Background agents, not teams | ✅ Experimental CLI teams |
+| **Task dependencies (blocked by)** | ✅ Guaranteed ordering | ✅ DAG waves via Beads | ✅ Blockers + execution locks | ❌ | ✅ Team task deps, no UI |
+| **Review workflow** | ✅ Agents review each other + human review UI | ⚠️ Refinery merge queue | ✅ Approvals + governance | ⚠️ PR/BugBot only | ✅ Team review, no UI |
+| **Zero setup** | ✅ Guided runtime setup | ❌ Go/Git/Dolt/Beads/tmux | ⚠️ `npx` + embedded Postgres | ✅ | ⚠️ CLI + env flag |
+| **Kanban board** | ✅ 5 columns, real-time | ❌ Dashboard, not Kanban | ✅ 7 columns, drag-and-drop | ❌ | ❌ |
+| **Execution log viewer** | ✅ Tool calls, reasoning, timeline | ⚠️ Feed, OTEL, dashboard | ✅ Run transcripts + ledger | ⚠️ Agent chat + terminal | ❌ |
+| **Live processes** | ✅ View, stop, open URLs in browser | ⚠️ Agent health dashboard | ⚠️ Manual services + previews | ⚠️ Native terminal only | ❌ |
+| **Per-task code review** | ✅ Accept / reject / comment | ⚠️ Merge queue, no diff UI | ⚠️ PR/work products, no inline diff | ✅ BugBot on PRs | ❌ |
+| **Flexible autonomy** | ✅ Per-action approvals + notifications | ✅ Gates, escalation, recovery | ✅ Board approvals, pause, terminate | ⚠️ BG agents auto-run commands | ✅ Permissions + hooks |
+| **Git worktree isolation** | ✅ Optional | ✅ Core primitive | ✅ Worktrees / branches | ⚠️ Background branches/VMs | ⚠️ Manual worktrees |
+| **Multi-agent backend** | ✅ Claude, Codex + OpenCode teammates | ✅ Claude, Codex, Gemini, Copilot + more | ✅ BYO agents: Claude, Codex, Cursor/OpenCode, HTTP | ⚠️ Multi-model agents, no team backend | ⚠️ Claude-only experimental teams |
+| **Org chart / governance** | ⚠️ Roles + approvals, no org chart | ⚠️ Roles + escalation | ✅ Org chart + board governance | ⚠️ Team admin only | ❌ |
+| **Budget controls** | ⚠️ Cost/token visibility, no hard caps | ⚠️ Cost tiers + digest, no hard caps | ✅ Per-agent budgets + hard stops | ⚠️ Usage + BG spend limits | ⚠️ `/cost` + workspace limits |
+| **Price** | **Free OSS UI**, provider access needed | Free OSS, runtime plans needed | Free OSS, self-hosted + infra | Free + paid usage | Claude plan or API usage |
+
+Fact sources checked on May 5, 2026: [detailed research notes](docs/research/gastown-paperclip-comparison-2026-05-05.md), [Gastown README](https://github.com/gastownhall/gastown), [Gastown provider guide](https://github.com/gastownhall/gastown/blob/main/docs/agent-provider-integration.md), [Gastown scheduler](https://github.com/gastownhall/gastown/blob/main/docs/design/scheduler.md), [Paperclip README](https://github.com/paperclipai/paperclip), [Paperclip adapters](https://github.com/paperclipai/paperclip/blob/master/docs/adapters/overview.md), [Paperclip budgets](https://github.com/paperclipai/paperclip/blob/master/docs/guides/board-operator/costs-and-budgets.md), [Paperclip runtime services](https://github.com/paperclipai/paperclip/blob/master/docs/guides/board-operator/execution-workspaces-and-runtime-services.md), [Paperclip Kanban source](https://github.com/paperclipai/paperclip/blob/master/ui/src/components/KanbanBoard.tsx), [Cursor Background Agents](https://docs.cursor.com/en/background-agents), [Cursor Diffs & Review](https://docs.cursor.com/en/agent/review), [Cursor Bugbot](https://docs.cursor.com/en/bugbot), [Cursor pricing](https://docs.cursor.com/en/account/usage), [Claude Code agent teams](https://code.claude.com/docs/en/agent-teams), [Claude Code subagents](https://code.claude.com/docs/en/sub-agents), [Claude Code workflows](https://code.claude.com/docs/en/common-workflows), [Claude Code costs](https://code.claude.com/docs/en/costs), [Claude pricing](https://claude.com/pricing).
 
 ---
 
diff --git a/docs/research/gastown-paperclip-comparison-2026-05-05.md b/docs/research/gastown-paperclip-comparison-2026-05-05.md
new file mode 100644
index 00000000..1570e5a3
--- /dev/null
+++ b/docs/research/gastown-paperclip-comparison-2026-05-05.md
@@ -0,0 +1,162 @@
+# Gastown и Paperclip comparison для лендинга и README
+
+> Дата проверки: 2026-05-05
+> Цель: публичная таблица `Agent Teams | Gastown | Paperclip | Cursor | Claude Code CLI` без угадываний по конкурентам.
+> Метод: первичные источники, `gh repo view`, official README/docs/releases, локальные клоны `gastownhall/gastown` и `paperclipai/paperclip`.
+
+## Snapshot
+
+| Проект | Позиционирование | Статус на 2026-05-05 | Лицензия |
+|---|---|---:|---|
+| **Gastown** | multi-agent workspace manager для coding agents | `14,962★`, latest `v1.0.1` от `2026-04-25`, push `2026-05-03` | MIT |
+| **Paperclip** | control plane для autonomous AI companies | `62,668★`, latest `v2026.428.0` от `2026-04-28`, push `2026-05-05` | MIT |
+
+## Что важно для публичного сравнения
+
+### Gastown
+
+Сильные факты:
+
+- README позиционирует Gastown как workspace manager для Claude Code, GitHub Copilot, Codex, Gemini и других AI agents.
+- Есть built-in mailboxes, identities, handoffs, Beads ledger, git-backed hooks, convoys, Witness/Deacon watchdog, Refinery merge queue.
+- Dependencies сильнее, чем просто "convoys": convoy stage/launch строит DAG, считает waves, запускает Wave 1, а последующие waves daemon dispatches автоматически.
+- Есть `gt feed`, web dashboard, OpenTelemetry events/metrics, `gt costs record`, cost tier presets, daily cost digest и context-budget guard.
+- Agent provider story шире README: provider guide прямо перечисляет Claude, Gemini, Codex, Cursor, AMP, OpenCode, Copilot и loose-coupled tmux integration tiers.
+- Scheduler сильнее простого queue: `scheduler.max_polecats` включает deferred dispatch, capacity governor, pause/resume, queued sling context beads и daemon dispatch cycle.
+
+Ограничения:
+
+- Это не Kanban product: README описывает dashboard как обзор agents, convoys, hooks, queues, issues, escalations.
+- Нет явного built-in editor, hunk-level review или task attachment workflow как продуктовой возможности.
+- Бюджеты не равны Paperclip budgets: у Gastown есть cost tiers/logging/guards, но я не нашёл hard monthly caps с automatic pause/cancel.
+
+Публичная оценка:
+
+- `Task dependencies` - `✅ Beads DAG waves`
+- `Kanban board` - `❌ Dashboard, not Kanban`
+- `Per-task code review` - `⚠️ Merge queue, no diff UI`
+- `Budget controls` - `⚠️ Cost tiers + digest, no hard caps`
+
+### Paperclip
+
+Сильные факты:
+
+- README: Node.js server + React UI that orchestrates a team of AI agents. Works with OpenClaw, Claude Code, Codex, Cursor, Bash, HTTP.
+- README прямо заявляет org charts, budgets, governance, goal alignment, agent coordination.
+- README under the hood: issues have company/project/goal/parent links, atomic checkout, execution locks, blocker dependencies, comments, documents, attachments, work products.
+- Budget section: scoped budget policies with warning thresholds and hard stops, overspend pauses agents and cancels queued work automatically.
+- `ui/src/components/KanbanBoard.tsx` содержит `backlog`, `todo`, `in_progress`, `in_review`, `blocked`, `done`, `cancelled` и drag-and-drop через `@dnd-kit`.
+- `packages/shared/src/validators/work-product.ts` содержит work products: `preview_url`, `runtime_service`, `pull_request`, `branch`, `commit`, `artifact`, `document`; statuses include `ready_for_review`, `approved`, `changes_requested`, `merged`.
+- Architecture docs: built-in stable adapters listed as Claude Local, Codex Local, Process, HTTP; adapter overview also lists OpenCode, Cursor, OpenClaw Gateway, Pi, Hermes and experimental Gemini.
+- Execution policy docs: runtime enforces comment-required, review and approval stages; decisions are audited; reviewers/approvers can be agents or users.
+- Workspace/runtime docs: services and jobs are manually controlled from UI, execution workspaces isolate checkout/branch/runtime state, and services are not auto-started by issue execution.
+- Issue API docs: @mentions in comments trigger heartbeats and issue-thread interactions can request confirmations, ask questions, or suggest tasks through UI cards.
+
+Ограничения:
+
+- README явно говорит: Paperclip is not a code review tool. It orchestrates work, not pull requests.
+- Нет hunk-level accept/reject UI и нет built-in code editor уровня нашего workbench.
+- Это broader company/org control plane, а не coding cockpit.
+- Work products are real in code, but public roadmap still has "Artifacts & Work Products" as future work, so public `Per-task code review` should stay `⚠️`, not `✅`.
+
+Публичная оценка:
+
+- `Kanban board` - `✅ 7 columns, drag-and-drop`
+- `Per-task code review` - `⚠️ PR/work products, no inline diff`
+- `Hunk-level review` - `❌ Bring your own review`
+- `Budget controls` - `✅ Per-agent budgets + hard stops`
+
+## Маркетинговая позиция
+
+Честный framing для нас:
+
+- Против Gastown продаём не "мы мощнее как orchestration OS", а **coding-team workbench**: review, logs, editor, live processes, attachments, operator UX.
+- Против Paperclip продаём не "у нас больше governance", а **agentic IDE / coding cockpit**: hunk-level review, task-scoped developer logs, built-in editor, live process controls.
+- `Multi-agent backend` у нас больше не надо показывать как "In development": публично корректнее писать `Claude, Codex + OpenCode teammates`.
+
+## Второй глубокий проход
+
+Что поменял после повторной проверки:
+
+- Убрал Gemini из нашей публичной строки. Теперь у нас: `Claude, Codex + OpenCode teammates`.
+- Paperclip `Agent-to-agent messaging` точнее как `Comments + @mentions`, а не как полноценный mailbox. @mentions будят агентов, но это не прямой peer mailbox как у нас или Gastown.
+- Paperclip `Execution log viewer` усилил до `Run transcripts + ledger`: в UI есть `RunTranscriptView` и `IssueRunLedger`, а architecture docs описывают capture stdout, cost/session state и run records.
+- Paperclip `Live processes` уточнил до `Manual services + previews`: runtime services/jobs есть в UI, но docs прямо говорят, что issue execution не стартует и не стопает их автоматически.
+- Gastown `Session analysis` уточнил до `Session recall, feed, OTEL`: это сильная ops/observability модель, но не наш task-scoped analysis cockpit.
+- Gastown `Budget controls` уточнил до `Cost tiers + digest, no hard caps`: есть model tiering, `gt costs record`, daily digest bead, но не нашёл Paperclip-style hard budget pause.
+
+Вывод не поменялся: публично выгоднее и честнее продавать наш **coding workbench** - hunk review, task logs, editor, live processes, attachments, team UI. Не надо притворяться, что мы глубже Gastown как orchestration OS или глубже Paperclip как governance/budget company control plane.
+
+## Третий глубокий проход по тексту таблицы
+
+Что уточнил после проверки README/docs/source:
+
+- Paperclip `Cross-team communication` заменил на `Company-scoped org work`: README говорит про multi-company isolation и org charts, но не про свободную коммуникацию между независимыми компаниями.
+- Paperclip `Agent-to-agent messaging` заменил на `Comments + @mentions`: docs называют comments primary channel, а @mentions будят агента через heartbeat. Это не peer mailbox.
+- Gastown `Linked tasks` заменил на `Beads deps + convoys`: source сильнее, чем просто "convoys", потому что stage/launch строит DAG по dependency edges.
+- Gastown `Session analysis` заменил на `Session recall, feed, OTEL`: публично понятнее, чем термин `Seance`, но всё ещё отражает session discovery, TUI feed и OpenTelemetry.
+- Gastown `Full autonomy` заменил на `Mayor, convoys, recovery`: не спорим с их глубокой autonomy моделью, но не перегружаем публичную таблицу внутренними health терминами.
+- Paperclip `Zero setup` заменил на `npx + embedded Postgres`: quickstart реально идёт через `npx paperclipai onboard --yes`, и docs говорят, что embedded PostgreSQL создаётся автоматически.
+- Paperclip `Price` уточнил до `OSS, self-hosted + infra`: честнее, потому что Paperclip account не нужен, но self-hosting и agent/provider runtime costs всё равно остаются.
+
+## Четвёртый глубокий проход: Claude Code CLI и Cursor
+
+Что поменял после проверки официальных docs Cursor и Claude Code:
+
+- Claude Code CLI нельзя честно держать как слабый baseline: официальные `agent teams` уже дают direct teammate messaging, shared task list, mailbox, task dependencies, team review scenarios и hooks for quality gates. Поэтому я поднял Claude CLI в строках `Agent-to-agent messaging`, `Linked tasks`, `Full autonomy`, `Task dependencies`, `Review workflow`.
+- Но Claude Code agent teams всё ещё experimental, disabled by default, требуют `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS`, работают в CLI/terminal display modes и имеют known limitations. Поэтому наш selling point против Claude CLI - не "у них нет teams", а **у них нет полноценного product UI/cockpit**: kanban, task attachments, hunk review UI, task logs, live process section, cross-team UI, cost/context dashboard.
+- Claude Code `Multi-agent backend` теперь точнее как `Claude-only experimental teams`, а не просто `Claude-only teams`.
+- Claude Code `Git worktree isolation` точнее как `Manual worktrees`: official workflows рекомендуют git worktree для параллельных sessions, но это не автоматическая продуктовая isolation как у нас/Gastown/Paperclip.
+- Cursor `Full autonomy` точнее как `Background agents, not teams`: background agents сильные, но это не team backend with shared task list/mailbox.
+- Cursor `Execution log viewer` и `Live processes` понижены до `⚠️`: есть agent chat/terminal/background terminals, но нет task-scoped execution timeline и process URL section как у нас.
+- Cursor `Flexible autonomy` понижен до `⚠️`: official background-agent security docs прямо говорят, что background agents auto-run terminal commands, в отличие от foreground approvals.
+- Cursor `Git worktree isolation` понижен до `⚠️ Background branches/VMs`: docs говорят про isolated remote machines and separate branches, не про встроенный local worktree strategy.
+- Cursor `Price` заменил с `$0-$200/mo` на `Free + paid usage`, потому что pricing/usage docs теперь акцентируют included API usage, usage dashboard, Bugbot pricing и background-agent spend limits.
+
+Итоговая позиция после четвёртого прохода:
+
+- Против Claude Code CLI мы не врём, что у него нет multi-agent primitives. У него они есть, но experimental and CLI-first. Мы продаём **операторский UI и review/workbench layer**.
+- Против Cursor мы не врём, что у него нет background agents или review. У него есть сильный IDE/PR story. Мы продаём **team/task orchestration, task-scoped logs, cross-agent workflow and review cockpit**.
+
+## Scores
+
+| Критерий | Agent Teams | Gastown | Paperclip |
+|---|---:|---:|---:|
+| Coding cockpit | **9.2** | 5.6 | 6.8 |
+| Orchestration depth | 7.6 | **9.2** | 8.8 |
+| Governance / budget control | 6.7 | 6.0 | **9.4** |
+| Review UX | **9.3** | 5.8 | 6.2 |
+| Setup simplicity | **8.4** | 4.7 | 7.1 |
+
+Вывод по 10-балльной уверенности:
+
+- Agent Teams против Gastown: 🎯 9   🛡️ 8   🧠 5 - сравнение честное, потому что мы отдаём Gastown orchestration depth, но забираем workbench.
+- Agent Teams против Paperclip: 🎯 9   🛡️ 8   🧠 6 - сравнение честное, потому что Paperclip сильнее в budgets/governance, но слабее как code-review/editor cockpit.
+- Agent Teams против Cursor: 🎯 9   🛡️ 9   🧠 5 - сравнение честное, потому что Cursor сильнее как IDE, но не как multi-agent team/task OS.
+- Agent Teams против Claude Code CLI: 🎯 9   🛡️ 9   🧠 6 - сравнение честное только после повышения Claude CLI по official agent teams primitives; наш выигрыш теперь UI/workbench, не наличие базовой team механики.
+- Публичная таблица после Claude/Cursor прохода: 🎯 9   🛡️ 9   🧠 5 - около 55 строк изменения в README/landing плюс локали.
+
+## Источники
+
+- Gastown repo: <https://github.com/gastownhall/gastown>
+- Gastown v1.0.1: <https://github.com/gastownhall/gastown/releases/tag/v1.0.1>
+- Gastown convoy implementation: <https://github.com/gastownhall/gastown/blob/main/internal/cmd/convoy_launch.go>
+- Gastown convoy skill docs: <https://github.com/gastownhall/gastown/tree/main/docs/skills/convoy>
+- Gastown provider guide: <https://github.com/gastownhall/gastown/blob/main/docs/agent-provider-integration.md>
+- Gastown scheduler docs: <https://github.com/gastownhall/gastown/blob/main/docs/design/scheduler.md>
+- Paperclip repo: <https://github.com/paperclipai/paperclip>
+- Paperclip v2026.428.0: <https://github.com/paperclipai/paperclip/releases/tag/v2026.428.0>
+- Paperclip Kanban source: <https://github.com/paperclipai/paperclip/blob/master/ui/src/components/KanbanBoard.tsx>
+- Paperclip work products source: <https://github.com/paperclipai/paperclip/blob/master/packages/shared/src/validators/work-product.ts>
+- Paperclip architecture docs: <https://github.com/paperclipai/paperclip/blob/master/docs/start/architecture.md>
+- Paperclip execution policy docs: <https://github.com/paperclipai/paperclip/blob/master/docs/guides/execution-policy.md>
+- Paperclip costs and budgets docs: <https://github.com/paperclipai/paperclip/blob/master/docs/guides/board-operator/costs-and-budgets.md>
+- Paperclip runtime services docs: <https://github.com/paperclipai/paperclip/blob/master/docs/guides/board-operator/execution-workspaces-and-runtime-services.md>
+- Cursor Background Agents: <https://docs.cursor.com/en/background-agents>
+- Cursor Diffs & Review: <https://docs.cursor.com/en/agent/review>
+- Cursor Bugbot: <https://docs.cursor.com/en/bugbot>
+- Cursor usage/pricing: <https://docs.cursor.com/en/account/usage>
+- Claude Code agent teams: <https://code.claude.com/docs/en/agent-teams>
+- Claude Code subagents: <https://code.claude.com/docs/en/sub-agents>
+- Claude Code common workflows: <https://code.claude.com/docs/en/common-workflows>
+- Claude Code costs: <https://code.claude.com/docs/en/costs>
diff --git a/docs/research/real-competitors-comparison.md b/docs/research/real-competitors-comparison.md
index b966a535..26121cae 100644
--- a/docs/research/real-competitors-comparison.md
+++ b/docs/research/real-competitors-comparison.md
@@ -1,5 +1,7 @@
 # Реальные конкуренты для Comparison в README
 
+> ⚠️ Update 2026-05-05: публичная таблица README/landing теперь сравнивает нас с `Gastown` и `Paperclip`, а не с `Claude Code Agent Teams` и `GoClaw`. Актуальная research-опора: [gastown-paperclip-comparison-2026-05-05.md](gastown-paperclip-comparison-2026-05-05.md). Ниже оставлен старый broader draft как исторический контекст.
+
 > Дата проверки: 2026-04-13  
 > Статус: внутренний comparison draft  
 > Цель: заменить в нашем внутреннем thinking `Vibe Kanban` и `Aperant` на реальные ориентиры - `Gastown`, `Claude Code Agent Teams`, `GoClaw`
@@ -55,7 +57,7 @@
 | **Per-task code review** | ✅ Per-task diff review with accept / reject / comment flow | ❌ | ❌ | ⚠️ Task approval exists, but not inline code diff review |
 | **Flexible autonomy** | ✅ Granular approvals, notifications, autonomy controls | ✅ Strong human gates, escalation and intervention, mostly via CLI/TUI | ⚠️ Plan approval, hooks and permissions exist, but control plane is thin | ✅ Team settings, approval workflows, exec approval, task approval |
 | **Git worktree isolation** | ✅ Optional per-agent worktree strategy | ✅ Core architectural primitive | ⚠️ Manual worktrees exist in Claude Code, but not as the native team model | ❌ Not a core team isolation model |
-| **Multi-agent backend** | ⚠️ Claude is mature; Codex/Gemini plumbing exists in code but is still emerging as product surface | ✅ Claude Code, Codex, Gemini, Copilot and other runtimes | ❌ Claude-first only, models per teammate but no real multi-provider backend | ✅ 20+ providers including Claude CLI and ChatGPT OAuth |
+| **Multi-agent backend** | ⚠️ Claude is mature; Codex/OpenCode plumbing exists in code but is still emerging as product surface | ✅ Claude Code, Codex, Gemini, Copilot and other runtimes | ❌ Claude-first only, models per teammate but no real multi-provider backend | ✅ 20+ providers including Claude CLI and ChatGPT OAuth |
 | **Price** | Free OSS UI, but a Claude Code plan is still needed today | Free OSS, but you still pay for the underlying runtime plans/seats you use | Claude subscription | Free self-hosted OSS, but infra + provider/API/subscription costs remain |
 
 ## Самые важные выводы по matrix
@@ -435,7 +437,7 @@ GoClaw выигрывает у нас по:
 
 ## Места, где надо быть особенно честными про нас
 
-- `Multi-agent backend` у нас пока не так зрел, как это можно прочитать из одной строки README. В коде есть мосты и статусы для `Anthropic`, `Codex`, `Gemini`, но продуктово основной путь всё ещё Claude-first.
+- `Multi-agent backend` у нас пока не так зрел, как это можно прочитать из одной строки README. В коде есть мосты и статусы для `Anthropic`, `Codex`, `OpenCode`, но продуктово основной путь всё ещё Claude-first.
 - `Zero setup` у нас честно сильный именно для Claude Code path.
 - `Cross-team communication` у нас сильнее, чем у этих конкурентов, но cross-team attachments не выглядят как полностью общий happy path.
 
diff --git a/landing/components/sections/ComparisonSection.vue b/landing/components/sections/ComparisonSection.vue
index c9c1ce6a..9c22083e 100644
--- a/landing/components/sections/ComparisonSection.vue
+++ b/landing/components/sections/ComparisonSection.vue
@@ -10,8 +10,8 @@ interface CellValue {
 interface ComparisonRow {
   feature: string
   us: CellValue
-  vibeKanban: CellValue
-  aperant: CellValue
+  gastown: CellValue
+  paperclip: CellValue
   cursor: CellValue
   claudeCli: CellValue
 }
@@ -20,161 +20,177 @@ const rows = computed<ComparisonRow[]>(() => [
   {
     feature: t('comparison.features.crossTeam'),
     us: { status: 'yes' },
-    vibeKanban: { status: 'no' },
-    aperant: { status: 'no' },
+    gastown: { status: 'partial', note: 'Cross-rig coordination' },
+    paperclip: { status: 'partial', note: 'Company-scoped org work' },
     cursor: { status: 'na' },
     claudeCli: { status: 'no' },
   },
   {
     feature: t('comparison.features.agentMessaging'),
     us: { status: 'yes', note: 'Native real-time mailbox' },
-    vibeKanban: { status: 'no', note: 'Agents are independent' },
-    aperant: { status: 'no', note: 'Fixed pipeline' },
+    gastown: { status: 'yes', note: 'Mailboxes + handoffs' },
+    paperclip: { status: 'partial', note: 'Comments + @mentions' },
     cursor: { status: 'no' },
-    claudeCli: { status: 'partial', note: 'Built-in (no UI)' },
+    claudeCli: { status: 'yes', note: 'Team mailbox, no UI' },
   },
   {
     feature: t('comparison.features.linkedTasks'),
-    us: { status: 'yes', note: 'Cross-references in messages' },
-    vibeKanban: { status: 'partial', note: 'Subtasks only' },
-    aperant: { status: 'no' },
+    us: { status: 'yes', note: 'Cross-refs + dependencies' },
+    gastown: { status: 'partial', note: 'Beads deps + convoys' },
+    paperclip: { status: 'yes', note: 'Goals, parents, blockers' },
     cursor: { status: 'no' },
-    claudeCli: { status: 'no' },
+    claudeCli: { status: 'yes', note: 'Shared task list' },
   },
   {
     feature: t('comparison.features.sessionAnalysis'),
-    us: { status: 'yes', note: '6-category token tracking' },
-    vibeKanban: { status: 'no' },
-    aperant: { status: 'partial', note: 'Execution logs' },
+    us: { status: 'yes', note: 'Task logs + token tracking' },
+    gastown: { status: 'partial', note: 'Session recall, feed, OTEL' },
+    paperclip: { status: 'partial', note: 'Run transcripts + cost audit' },
     cursor: { status: 'no' },
-    claudeCli: { status: 'no' },
+    claudeCli: { status: 'partial', note: 'Usage command, no UI' },
   },
   {
     feature: t('comparison.features.taskAttachments'),
     us: { status: 'yes', note: 'Auto-attach, agents read & attach' },
-    vibeKanban: { status: 'no' },
-    aperant: { status: 'yes', note: 'Images + files' },
+    gastown: { status: 'no', note: 'Not task-level' },
+    paperclip: { status: 'yes', note: 'Docs, attachments, work products' },
     cursor: { status: 'partial', note: 'Chat session only' },
-    claudeCli: { status: 'no' },
+    claudeCli: { status: 'partial', note: 'Chat images only' },
   },
   {
     feature: t('comparison.features.hunkReview'),
     us: { status: 'yes', note: 'Accept / reject individual hunks' },
-    vibeKanban: { status: 'no' },
-    aperant: { status: 'no' },
+    gastown: { status: 'no' },
+    paperclip: { status: 'no', note: 'Bring your own review' },
     cursor: { status: 'yes' },
     claudeCli: { status: 'no' },
   },
   {
     feature: t('comparison.features.codeEditor'),
     us: { status: 'yes', note: 'With Git support' },
-    vibeKanban: { status: 'no' },
-    aperant: { status: 'no' },
+    gastown: { status: 'no' },
+    paperclip: { status: 'no', note: 'Control plane, not editor' },
     cursor: { status: 'yes', note: 'Full IDE' },
     claudeCli: { status: 'no' },
   },
   {
     feature: t('comparison.features.fullAutonomy'),
     us: { status: 'yes', note: 'Create, assign, review end-to-end' },
-    vibeKanban: { status: 'no', note: 'Human manages tasks' },
-    aperant: { status: 'no', note: 'Fixed pipeline' },
-    cursor: { status: 'partial', note: 'Isolated tasks only' },
-    claudeCli: { status: 'partial', note: 'No UI' },
+    gastown: { status: 'yes', note: 'Mayor, convoys, recovery' },
+    paperclip: { status: 'yes', note: 'Heartbeats + governance' },
+    cursor: { status: 'partial', note: 'Background agents, not teams' },
+    claudeCli: { status: 'yes', note: 'Experimental CLI teams' },
   },
   {
     feature: t('comparison.features.taskDeps'),
     us: { status: 'yes', note: 'Guaranteed ordering' },
-    vibeKanban: { status: 'no' },
-    aperant: { status: 'partial', note: 'Within plan only' },
+    gastown: { status: 'yes', note: 'DAG waves via Beads' },
+    paperclip: { status: 'yes', note: 'Blockers + execution locks' },
     cursor: { status: 'no' },
-    claudeCli: { status: 'partial', note: 'No UI, no notifications' },
+    claudeCli: { status: 'yes', note: 'Team task deps, no UI' },
   },
   {
     feature: t('comparison.features.reviewWorkflow'),
     us: { status: 'yes', note: 'Agents review each other' },
-    vibeKanban: { status: 'no' },
-    aperant: { status: 'partial', note: 'Auto QA pipeline' },
-    cursor: { status: 'no' },
-    claudeCli: { status: 'partial', note: 'No UI' },
+    gastown: { status: 'partial', note: 'Refinery merge queue' },
+    paperclip: { status: 'yes', note: 'Approvals + governance' },
+    cursor: { status: 'partial', note: 'PR/BugBot only' },
+    claudeCli: { status: 'yes', note: 'Team review, no UI' },
   },
   {
     feature: t('comparison.features.zeroSetup'),
-    us: { status: 'yes' },
-    vibeKanban: { status: 'no', note: 'Config required' },
-    aperant: { status: 'no', note: 'Config required' },
+    us: { status: 'yes', note: 'Guided runtime setup' },
+    gastown: { status: 'no', note: 'Go/Git/Dolt/Beads/tmux' },
+    paperclip: { status: 'partial', note: 'npx + embedded Postgres' },
     cursor: { status: 'yes' },
-    claudeCli: { status: 'partial', note: 'CLI install required' },
+    claudeCli: { status: 'partial', note: 'CLI + env flag' },
   },
   {
     feature: t('comparison.features.kanban'),
     us: { status: 'yes', note: '5 columns, real-time' },
-    vibeKanban: { status: 'yes' },
-    aperant: { status: 'yes', note: '6 columns (pipeline)' },
+    gastown: { status: 'no', note: 'Dashboard, not Kanban' },
+    paperclip: { status: 'yes', note: '7 columns, drag-and-drop' },
     cursor: { status: 'no' },
     claudeCli: { status: 'no' },
   },
   {
     feature: t('comparison.features.execLog'),
     us: { status: 'yes', note: 'Tool calls, reasoning, timeline' },
-    vibeKanban: { status: 'no' },
-    aperant: { status: 'yes', note: 'Phase-based logs' },
-    cursor: { status: 'yes' },
+    gastown: { status: 'partial', note: 'Feed, OTEL, dashboard' },
+    paperclip: { status: 'yes', note: 'Run transcripts + ledger' },
+    cursor: { status: 'partial', note: 'Agent chat + terminal' },
     claudeCli: { status: 'no' },
   },
   {
     feature: t('comparison.features.liveProcesses'),
     us: { status: 'yes', note: 'View, stop, open URLs' },
-    vibeKanban: { status: 'no' },
-    aperant: { status: 'no' },
-    cursor: { status: 'yes' },
+    gastown: { status: 'partial', note: 'Agent health dashboard' },
+    paperclip: { status: 'partial', note: 'Manual services + previews' },
+    cursor: { status: 'partial', note: 'Native terminal only' },
     claudeCli: { status: 'no' },
   },
   {
     feature: t('comparison.features.perTaskReview'),
     us: { status: 'yes', note: 'Accept / reject / comment' },
-    vibeKanban: { status: 'partial', note: 'PR-level only' },
-    aperant: { status: 'partial', note: 'File-level only' },
+    gastown: { status: 'partial', note: 'Merge queue, no diff UI' },
+    paperclip: { status: 'partial', note: 'PR/work products, no diff UI' },
     cursor: { status: 'yes', note: 'BugBot on PRs' },
     claudeCli: { status: 'no' },
   },
   {
     feature: t('comparison.features.flexAutonomy'),
-    us: { status: 'yes', note: 'Granular settings, notifications' },
-    vibeKanban: { status: 'no' },
-    aperant: { status: 'partial', note: 'Plan approval only' },
-    cursor: { status: 'yes' },
-    claudeCli: { status: 'yes' },
+    us: { status: 'yes', note: 'Per-action approvals + notifications' },
+    gastown: { status: 'yes', note: 'Gates, escalation, recovery' },
+    paperclip: { status: 'yes', note: 'Board approvals, pause, terminate' },
+    cursor: { status: 'partial', note: 'BG agents auto-run commands' },
+    claudeCli: { status: 'yes', note: 'Permissions + hooks' },
   },
   {
     feature: t('comparison.features.worktree'),
     us: { status: 'yes', note: 'Optional' },
-    vibeKanban: { status: 'partial', note: 'Mandatory' },
-    aperant: { status: 'partial', note: 'Mandatory' },
-    cursor: { status: 'yes' },
-    claudeCli: { status: 'yes' },
+    gastown: { status: 'yes', note: 'Core primitive' },
+    paperclip: { status: 'yes', note: 'Worktrees / branches' },
+    cursor: { status: 'partial', note: 'Background branches/VMs' },
+    claudeCli: { status: 'partial', note: 'Manual worktrees' },
   },
   {
     feature: t('comparison.features.multiAgent'),
-    us: { status: 'soon', note: 'In development', noteLink: 'https://github.com/Alishahryar1/free-claude-code' },
-    vibeKanban: { status: 'yes', note: '6+ agents' },
-    aperant: { status: 'yes', note: '11 providers' },
-    cursor: { status: 'yes', note: 'Multi-model' },
-    claudeCli: { status: 'na' },
+    us: { status: 'yes', note: 'Claude, Codex + OpenCode teammates' },
+    gastown: { status: 'yes', note: 'Claude, Codex, Gemini, Copilot + more' },
+    paperclip: { status: 'yes', note: 'BYO agents: Claude, Codex, Cursor/OpenCode, HTTP' },
+    cursor: { status: 'partial', note: 'Multi-model agents, no team backend' },
+    claudeCli: { status: 'partial', note: 'Claude-only experimental teams' },
+  },
+  {
+    feature: t('comparison.features.orgGovernance'),
+    us: { status: 'partial', note: 'Roles + approvals, no org chart' },
+    gastown: { status: 'partial', note: 'Roles + escalation' },
+    paperclip: { status: 'yes', note: 'Org chart + board governance' },
+    cursor: { status: 'partial', note: 'Team admin only' },
+    claudeCli: { status: 'no' },
+  },
+  {
+    feature: t('comparison.features.budgetControls'),
+    us: { status: 'partial', note: 'Cost/token visibility, no hard caps' },
+    gastown: { status: 'partial', note: 'Cost tiers + digest, no hard caps' },
+    paperclip: { status: 'yes', note: 'Per-agent budgets + hard stops' },
+    cursor: { status: 'partial', note: 'Usage + BG spend limits' },
+    claudeCli: { status: 'partial', note: '/cost + workspace limits' },
   },
   {
     feature: t('comparison.features.price'),
-    us: { status: 'free' },
-    vibeKanban: { status: 'text', note: 'Free / $30/mo' },
-    aperant: { status: 'free' },
-    cursor: { status: 'text', note: '$0–$200/mo' },
-    claudeCli: { status: 'text', note: 'Claude subscription' },
+    us: { status: 'free', note: 'OSS, provider access needed' },
+    gastown: { status: 'free', note: 'OSS, runtime plans needed' },
+    paperclip: { status: 'free', note: 'OSS, self-hosted + infra' },
+    cursor: { status: 'text', note: 'Free + paid usage' },
+    claudeCli: { status: 'text', note: 'Claude plan or API usage' },
   },
 ])
 
 const competitors = [
   { key: 'us', name: 'Agent Teams', highlight: true },
-  { key: 'vibeKanban', name: 'Vibe Kanban' },
-  { key: 'aperant', name: 'Aperant' },
+  { key: 'gastown', name: 'Gastown' },
+  { key: 'paperclip', name: 'Paperclip' },
   { key: 'cursor', name: 'Cursor' },
   { key: 'claudeCli', name: 'Claude Code CLI' },
 ]
@@ -197,7 +213,7 @@ function getStatusIcon(status: string): string {
     case 'yes': return '\u2713'
     case 'no': return '\u2717'
     case 'partial': return '\u25D2'
-    case 'na': return '\u2014'
+    case 'na': return 'N/A'
     case 'free': return 'Free'
     case 'soon': return '\uD83D\uDCC5'
     default: return ''
diff --git a/landing/locales/ar.json b/landing/locales/ar.json
index 316ffad0..295142a1 100644
--- a/landing/locales/ar.json
+++ b/landing/locales/ar.json
@@ -83,6 +83,8 @@
       "flexAutonomy": "استقلالية مرنة",
       "worktree": "عزل Git worktree",
       "multiAgent": "خلفية متعددة الوكلاء",
+      "orgGovernance": "الهيكل التنظيمي / الحوكمة",
+      "budgetControls": "ضوابط الميزانية",
       "price": "السعر"
     }
   },
diff --git a/landing/locales/de.json b/landing/locales/de.json
index 7373d66b..76257315 100644
--- a/landing/locales/de.json
+++ b/landing/locales/de.json
@@ -83,6 +83,8 @@
       "flexAutonomy": "Flexible Autonomie",
       "worktree": "Git-Worktree-Isolation",
       "multiAgent": "Multi-Agenten-Backend",
+      "orgGovernance": "Organigramm / Governance",
+      "budgetControls": "Budgetkontrollen",
       "price": "Preis"
     }
   },
diff --git a/landing/locales/en.json b/landing/locales/en.json
index dc79746f..da3c26cb 100644
--- a/landing/locales/en.json
+++ b/landing/locales/en.json
@@ -83,6 +83,8 @@
       "flexAutonomy": "Flexible autonomy",
       "worktree": "Git worktree isolation",
       "multiAgent": "Multi-agent backend",
+      "orgGovernance": "Org chart / governance",
+      "budgetControls": "Budget controls",
       "price": "Price"
     }
   },
diff --git a/landing/locales/es.json b/landing/locales/es.json
index f911ff4f..5cb5c78d 100644
--- a/landing/locales/es.json
+++ b/landing/locales/es.json
@@ -83,6 +83,8 @@
       "flexAutonomy": "Autonomía flexible",
       "worktree": "Aislamiento Git worktree",
       "multiAgent": "Backend multi-agente",
+      "orgGovernance": "Organigrama / gobernanza",
+      "budgetControls": "Controles de presupuesto",
       "price": "Precio"
     }
   },
diff --git a/landing/locales/fr.json b/landing/locales/fr.json
index 0af848a1..3566ed88 100644
--- a/landing/locales/fr.json
+++ b/landing/locales/fr.json
@@ -83,6 +83,8 @@
       "flexAutonomy": "Autonomie flexible",
       "worktree": "Isolation Git worktree",
       "multiAgent": "Backend multi-agents",
+      "orgGovernance": "Organigramme / gouvernance",
+      "budgetControls": "Contrôle budgétaire",
       "price": "Prix"
     }
   },
diff --git a/landing/locales/hi.json b/landing/locales/hi.json
index 4fb4ad8a..ff06e28a 100644
--- a/landing/locales/hi.json
+++ b/landing/locales/hi.json
@@ -83,6 +83,8 @@
       "flexAutonomy": "लचीली स्वायत्तता",
       "worktree": "Git worktree आइसोलेशन",
       "multiAgent": "मल्टी-एजेंट बैकएंड",
+      "orgGovernance": "ऑर्ग चार्ट / गवर्नेंस",
+      "budgetControls": "बजट नियंत्रण",
       "price": "कीमत"
     }
   },
diff --git a/landing/locales/ja.json b/landing/locales/ja.json
index efe26554..04b1c92c 100644
--- a/landing/locales/ja.json
+++ b/landing/locales/ja.json
@@ -83,6 +83,8 @@
       "flexAutonomy": "柔軟な自律性",
       "worktree": "Git worktree分離",
       "multiAgent": "マルチエージェントバックエンド",
+      "orgGovernance": "組織図 / ガバナンス",
+      "budgetControls": "予算管理",
       "price": "価格"
     }
   },
diff --git a/landing/locales/pt.json b/landing/locales/pt.json
index ba1c26ca..f13006b6 100644
--- a/landing/locales/pt.json
+++ b/landing/locales/pt.json
@@ -83,6 +83,8 @@
       "flexAutonomy": "Autonomia flexível",
       "worktree": "Isolamento Git worktree",
       "multiAgent": "Backend multi-agente",
+      "orgGovernance": "Organograma / governança",
+      "budgetControls": "Controles de orçamento",
       "price": "Preço"
     }
   },
diff --git a/landing/locales/ru.json b/landing/locales/ru.json
index f57612bf..bde5b30e 100644
--- a/landing/locales/ru.json
+++ b/landing/locales/ru.json
@@ -83,6 +83,8 @@
       "flexAutonomy": "Гибкая автономность",
       "worktree": "Изоляция Git worktree",
       "multiAgent": "Мультиагентный бэкенд",
+      "orgGovernance": "Оргструктура / управление",
+      "budgetControls": "Бюджетные лимиты",
       "price": "Цена"
     }
   },
diff --git a/landing/locales/zh.json b/landing/locales/zh.json
index bd56e97f..712972a3 100644
--- a/landing/locales/zh.json
+++ b/landing/locales/zh.json
@@ -83,6 +83,8 @@
       "flexAutonomy": "灵活自主",
       "worktree": "Git worktree 隔离",
       "multiAgent": "多智能体后端",
+      "orgGovernance": "组织架构 / 治理",
+      "budgetControls": "预算控制",
       "price": "价格"
     }
   },
diff --git a/runtime.lock.json b/runtime.lock.json
index 481af2ff..d7894dfa 100644
--- a/runtime.lock.json
+++ b/runtime.lock.json
@@ -1,27 +1,27 @@
 {
-  "version": "0.0.17",
-  "sourceRef": "v0.0.17",
+  "version": "0.0.18",
+  "sourceRef": "v0.0.18",
   "sourceRepository": "777genius/agent_teams_orchestrator",
   "releaseRepository": "777genius/claude_agent_teams_ui",
   "releaseTag": "v1.2.0",
   "assets": {
     "darwin-arm64": {
-      "file": "agent-teams-runtime-darwin-arm64-v0.0.17.tar.gz",
+      "file": "agent-teams-runtime-darwin-arm64-v0.0.18.tar.gz",
       "archiveKind": "tar.gz",
       "binaryName": "claude-multimodel"
     },
     "darwin-x64": {
-      "file": "agent-teams-runtime-darwin-x64-v0.0.17.tar.gz",
+      "file": "agent-teams-runtime-darwin-x64-v0.0.18.tar.gz",
       "archiveKind": "tar.gz",
       "binaryName": "claude-multimodel"
     },
     "linux-x64": {
-      "file": "agent-teams-runtime-linux-x64-v0.0.17.tar.gz",
+      "file": "agent-teams-runtime-linux-x64-v0.0.18.tar.gz",
       "archiveKind": "tar.gz",
       "binaryName": "claude-multimodel"
     },
     "win32-x64": {
-      "file": "agent-teams-runtime-win32-x64-v0.0.17.zip",
+      "file": "agent-teams-runtime-win32-x64-v0.0.18.zip",
       "archiveKind": "zip",
       "binaryName": "claude-multimodel.exe"
     }
diff --git a/test/main/services/team/MemberWorkSyncClaudeStopHook.live.test.ts b/test/main/services/team/MemberWorkSyncClaudeStopHook.live.test.ts
index 21d8d7c7..5cb1e78e 100644
--- a/test/main/services/team/MemberWorkSyncClaudeStopHook.live.test.ts
+++ b/test/main/services/team/MemberWorkSyncClaudeStopHook.live.test.ts
@@ -40,9 +40,11 @@ vi.mock('../../../../src/main/services/infrastructure/NotificationManager', () =
   },
 }));
 
+const allowConnectedClaudeAccount =
+  process.env.MEMBER_WORK_SYNC_CLAUDE_ALLOW_CONNECTED_ACCOUNT === '1';
 const liveDescribe =
   process.env.MEMBER_WORK_SYNC_CLAUDE_STOP_HOOK_LIVE === '1' &&
-  Boolean(process.env.ANTHROPIC_API_KEY?.trim())
+  (hasLiveAnthropicApiKey() || allowConnectedClaudeAccount)
     ? describe
     : describe.skip;
 
@@ -103,9 +105,7 @@ liveDescribe('Member work sync Claude Stop hook live e2e', () => {
   beforeEach(async () => {
     tempDir = await fs.mkdtemp(path.join(os.tmpdir(), 'member-work-sync-claude-stop-live-'));
     tempClaudeRoot = path.join(tempDir, '.claude');
-    tempHome = path.join(tempDir, 'home');
     await fs.mkdir(tempClaudeRoot, { recursive: true });
-    await fs.mkdir(tempHome, { recursive: true });
     setClaudeBasePathOverride(tempClaudeRoot);
 
     previousCliPath = process.env.CLAUDE_AGENT_TEAMS_ORCHESTRATOR_CLI_PATH;
@@ -117,6 +117,12 @@ liveDescribe('Member work sync Claude Stop hook live e2e', () => {
     previousHistFile = process.env.HISTFILE;
     previousUserProfile = process.env.USERPROFILE;
 
+    const shouldUseConnectedAccountHome = allowConnectedClaudeAccount && !hasLiveAnthropicApiKey();
+    tempHome = shouldUseConnectedAccountHome
+      ? resolveConnectedClaudeHome(previousHome)
+      : path.join(tempDir, 'home');
+    await fs.mkdir(tempHome, { recursive: true });
+
     process.env.HOME = tempHome;
     process.env.HISTFILE = '/dev/null';
     process.env.USERPROFILE = tempHome;
@@ -531,3 +537,19 @@ async function cleanupScopedClaudeStopHookLiveTempDirs(): Promise<void> {
     }
   }
 }
+
+function hasLiveAnthropicApiKey(): boolean {
+  return Boolean(process.env.ANTHROPIC_API_KEY?.trim());
+}
+
+function resolveConnectedClaudeHome(previousHome: string | undefined): string {
+  const explicit = process.env.MEMBER_WORK_SYNC_CLAUDE_CONNECTED_HOME?.trim();
+  if (explicit) {
+    return path.resolve(explicit);
+  }
+  const previous = previousHome?.trim();
+  if (previous) {
+    return path.resolve(previous);
+  }
+  return os.userInfo().homedir;
+}
diff --git a/test/main/services/team/MemberWorkSyncCodex.live.test.ts b/test/main/services/team/MemberWorkSyncCodex.live.test.ts
index 0681b51c..88f7ef26 100644
--- a/test/main/services/team/MemberWorkSyncCodex.live.test.ts
+++ b/test/main/services/team/MemberWorkSyncCodex.live.test.ts
@@ -56,6 +56,7 @@ liveDescribe('Member work sync Codex live e2e', () => {
   let previousCliFlavor: string | undefined;
   let previousControlUrl: string | undefined;
   let previousCodexHome: string | undefined;
+  let previousCodexIgnoreUserConfig: string | undefined;
   let codexHomeDir: string;
   let ownsCodexHomeDir: boolean;
   let codexAccountFeature: {
@@ -100,6 +101,7 @@ liveDescribe('Member work sync Codex live e2e', () => {
     previousCliFlavor = process.env.CLAUDE_TEAM_CLI_FLAVOR;
     previousControlUrl = process.env.CLAUDE_TEAM_CONTROL_URL;
     previousCodexHome = process.env.CODEX_HOME;
+    previousCodexIgnoreUserConfig = process.env.CLAUDE_CODE_CODEX_NATIVE_IGNORE_USER_CONFIG;
 
     const shouldUseConnectedAccountHome = allowConnectedChatGptAccount && !hasLiveCodexApiKey();
     if (shouldUseConnectedAccountHome) {
@@ -117,6 +119,7 @@ liveDescribe('Member work sync Codex live e2e', () => {
       process.env.CLAUDE_AGENT_TEAMS_ORCHESTRATOR_CLI_PATH?.trim() || DEFAULT_ORCHESTRATOR_CLI;
     process.env.CLAUDE_TEAM_CLI_FLAVOR = 'agent_teams_orchestrator';
     process.env.CODEX_HOME = codexHomeDir;
+    process.env.CLAUDE_CODE_CODEX_NATIVE_IGNORE_USER_CONFIG = 'true';
 
     codexAccountFeature = null;
     providerConnectionService = null;
@@ -141,6 +144,7 @@ liveDescribe('Member work sync Codex live e2e', () => {
     restoreEnv('CLAUDE_TEAM_CLI_FLAVOR', previousCliFlavor);
     restoreEnv('CLAUDE_TEAM_CONTROL_URL', previousControlUrl);
     restoreEnv('CODEX_HOME', previousCodexHome);
+    restoreEnv('CLAUDE_CODE_CODEX_NATIVE_IGNORE_USER_CONFIG', previousCodexIgnoreUserConfig);
     setClaudeBasePathOverride(null);
     if (process.env.MEMBER_WORK_SYNC_CODEX_KEEP_TEMP === '1') {
       console.info(`[MemberWorkSyncCodex.live] preserved temp dir: ${tempDir}`);
@@ -254,7 +258,7 @@ liveDescribe('Member work sync Codex live e2e', () => {
           prompt: [
             'Keep launch work minimal.',
             'If you receive a task, follow task instructions exactly.',
-            'Before going idle with unfinished assigned work, call member_work_sync_status and member_work_sync_report.',
+            'Do not call member_work_sync_status until a task instruction or member_work_sync_nudge provides exact teamName, memberName, and controlUrl.',
           ].join(' '),
           members: [],
         },
@@ -292,19 +296,16 @@ liveDescribe('Member work sync Codex live e2e', () => {
         ].join('\n'),
       });
       feature.noteTeamChange({ type: 'task', teamName, taskId: task.id });
+
+      const preRelayStatus = await feature.refreshStatus({ teamName, memberName });
+      expect(preRelayStatus.memberName).toBe(memberName);
+      expect(preRelayStatus.providerId).toBe('codex');
+      expect(preRelayStatus.agenda.items.some((item) => item.taskId === task.id)).toBe(true);
+      expect(preRelayStatus.shadow?.wouldNudge).toBe(true);
+
       const relay = await activeService.relayInboxFileToLiveRecipient(teamName, memberName);
       expect(relay.relayed).toBeGreaterThan(0);
 
-      await waitUntil(async () => {
-        const status = await feature!.getStatus({ teamName: teamName!, memberName });
-        return (
-          status.memberName === memberName &&
-          status.providerId === 'codex' &&
-          status.agenda.items.some((item) => item.taskId === task.id) &&
-          status.shadow?.wouldNudge === true
-        );
-      }, 30_000);
-
       await waitUntil(async () => {
         const fatalRuntimeMessage = await readFatalRuntimeMessage(teamName!);
         if (fatalRuntimeMessage) {