Using the WebUI — administration (「管理」)
Any screen not listed here does not exist. Headings quote the screen's own wording.
Part of the settings guide. Other topics:
conversations, the input box, and sign-in: references/webui.md;
documents, files, tables, memory, channels, and apps: references/webui-workspace.md;
editing a file in place (「編集」): references/webui-file-edit.md;
SIROE pages, blocks, and boards: references/webui-siroe.md;
settings, administration, models, and triggers: references/webui-settings.md;
external service connections: references/webui-connections.md;
/office: references/webui-office.md;
/meetings: references/webui-meetings.md;
/calendar: references/webui-calendar.md.
「管理」
| Where | What it configures |
|---|---|
/settings/workspace |
「基本情報」. The workspace name and ID (include it in inquiries). Admins and managers can change the name. 「ワークスペースの削除」 appears only for admins and deletes all members' data (recoverable for 90 days by contacting the operator); a personal workspace can be deleted only by the person who created it, even if others are admins there |
/settings/members |
「メンバー」. 3 tabs (?tab=members / groups / sso). 「メンバー」 is participants and invitations. Roles are オーナー / 管理者 / マネージャー / メンバー. Inviting, changing roles, and removing are done by admins and managers; only admins can make someone an admin, invite as an admin, or change or remove an admin (a manager's role choices stop at マネージャー; nobody on their own row). A large workspace lists the first 100 people; 「さらに 100 人表示」 at the bottom shows more (searching narrows the whole list). Check rows to change roles, add to a group, or remove several at once — rows you cannot change are skipped, and the confirmation says how many; a group's 管理担当 sees the checkboxes too, for adding to their groups. An invitation can name several groups (「グループ(任意)」); the invited person joins all of them on accepting. The オーナー is the single person who created the workspace, with the same available operations as an admin (the オーナー's row cannot be changed on screen). Invitations arrive by email (if not, hand over the row's link; max 100 per 24 hours). From a row's 「サインイン中の端末」, only admins can end that person's sign-ins one device at a time or all at once (it is tied to the person, so their other workspaces are signed out too); the オーナー's devices can be seen and ended only by the オーナー themself. From 「外部チャットの紐付け」, only admins can 「解除」 that person's linked external chat accounts (external ids are not shown); calls from it then stop acting as them. They can relink; removing the member drops every link. 「グループ」 groups people by department (up to 2 levels of nesting). Anyone can click a group's row to see who is in it; the row says 「あなたが所属」 or 「管理担当」 for the viewer. Rename / move / delete are in the row's 「…」 menu. Creating, deleting, and moving parents is done by admins and managers. Each group can have a 「管理担当」 — for that group and below only, they can add/remove people, rename, and appoint/remove. A 管理担当 can also invite from 「招待する」 on the 「メンバー」 tab — only with the メンバー role, and choosing at least one of the groups they manage is required; their 「保留中の招待」 shows only invitations into those groups. Used as a knowledge sharing target. 「シングルサインオン」 is settings for signing in with a company account (OIDC / SAML) (admins only). Currently being validated; to use it, contact your representative. Even once registered, it is unusable until verified by a DNS TXT record |
/settings/permissions |
「許可設定」. 4 tabs (?tab=network / access / execution / suggestions). 「外部との通信」 — accessible external domains (destinations connected so far, with origin; 「すべての宛先を許可する」 (on by default: secret-free traffic may go anywhere); admins and managers switch it, add hosts, and block rows. Blocking covers subdomains and overrides approvals) / hosts where browser operation is prohibited (viewing, typing, and in-page scripts are all blocked there, in the agent's browser and on connected devices; register subdomains individually) / 「取り込んだ文章の確認」 (mode 「記録するだけ」 / 「取り込みを停止する」, plus a switch per source: Web incl. browser operation, knowledge/files/DB, MCP, conversations/memory, skills, command output; off = no judging calls or records). 「接続元」 — connection source restrictions / |
| 端末記憶の許可 (organization accounts only; on by default, existing ones unchanged. On lets members tick | |
| 「この端末を 30 日間覚える」 on the second sign-in step and skip the emailed code on that device | |
| for 30 days. Turning it off applies at once, even to remembered ones. Unusable while | |
any other organization of the member has it off). 「実行と承認」 — app execution permissions / per-tool handling (which operations raise an approval card; read rows offer only 「既定に従う」/「禁止」; a search box filters rows by name, 「変更したものだけ」 shows only rows set explicitly, and each group heading shows how many were changed; the circular-arrow button beside a set row returns it to 「既定に従う」). 「自動処理」 — 今週の提案 (weekly, proposes turning repeated work into skills or 定期実行; draws on others' conversations but shows only rewritten summaries) / 記憶の週次提案 (weekly, lists facts worth remembering from the person's own conversations last week on their first screen; nothing enters memory until pressed) / 会話タイトルの自動生成 (AI titles each new chat once; off keeps its first words) / ミーティングの記録の保持期間 (90 / 180 (default) / 365 日 or 無期限; after it, recordings, audio, and transcripts are removed and minutes stay — see references/webui-meetings-records.md). Admins and managers change these; per-tool handling and prohibited hosts: admins only. Current values are readable by everyone, except the 2 weekly items (admins and managers only) |
|
/settings/usage |
「利用状況」 (the numbers are readable by everyone; only admins can set limits and enable the balance). 3 tabs — 「概要」「内訳」「残高と上限」 — showing credit consumption and run time per person and period. The breakdown also has two platform rows nobody triggers directly — 「データベース処理」 (data work behind conversations, settings, and records; counted hourly, about an hour late) and 「保管」 (stored files and records; counted daily). Both consume credits like the other rows. Totals count only what we bill — 持ち込み鍵 (own API key) use is on a separate line, outside billing and limits. Record retention periods are also stated here — conversations not used for 180 days are deleted automatically with their content (pinned ones, ones open to the public, and team ones remain). Trigger run records, settled approval requests, and child-agent run logs also go after 180 days (the parent exchange remains until the conversation's expiry). Reviewed compliance findings are deleted with the conversation (unreviewed ones stay until reviewed) |
/settings/subscription |
「契約情報」. 3 tabs. 「契約プラン」 is applying for, changing, and cancelling credits (admin AND organization account; individuals wait for the operator). If an annual contract proposal has arrived, it is accepted on the same tab. The amount is base fee + credits, 本体価格(税抜). Amount changes take effect from the 1st of next month; if this month falls short, 「クレジットの追加」 buys more on the spot (in steps of 10, monthly billing only — not available on annual contracts; expires on the 1st of next month; charged on the next invoice). 「支払い」 is the next scheduled invoice (including additions) and purchase history. 「支払い情報」 is the addressee, delivery addresses (up to 5), and address |
/settings/compliance |
「コンプライアンス」. Checks daily whether what members sent to the AI violates defined policies (the list is readable by everyone, but utterance excerpts, detection reasons, and conversation titles and full text are for admins and managers only; others see 「権限がないため非表示」). 4 tabs — 「見つかったもの」「規程と巡回」「メンバーの会話」「監査ログ」 (?tab=findings / rules / oversight / audit); the sidebar has no own rows for the last two. 「規程と巡回」 shows for admins only, 「メンバーの会話」 for admins and managers only. Off by default (admins turn it on), and the daily patrol covers only utterances made after it was enabled. While on, members can read this under 「この職場の決まり」 in /settings/account (nothing appears in the chat input). A new finding in a one-on-one conversation sends that person a note in the conversation, a push, and an email naming only the policy; admins and managers get a push and an email with only the count. Besides turning the provided policies ON/OFF, your own company policies can be added as text. Only admins and managers can resolve found rows as 「違反と確認」 or 「誤検知」 (hovering either button says what it does). Neither notifies the speaker, and 「誤検知」 does not feed back into later checks — if the same false positive keeps coming, revise the policy text (or turn the policy off). Names of the workspace's own members are not flagged as outside people's personal information, and a question sent to someone's 影武者 is checked as the asker's utterance, not the owner's. 「メンバーの会話」 reads the conversations of people in lower roles than yourself (only strictly lower-ranked people; peers such as admin–admin cannot). Switched on/off on that tab (off by default; admins only). Members can see whether they may be read under 「この職場の決まり」 in /settings/account. Only that person's one-on-one conversations appear; channels, public windows, and deleted ones do not. Both opening the list and opening full text are recorded in the audit log. 「監査ログ」 (the 3 columns operation, actor, and date/time are readable by everyone. The 「詳細」 column is admins only; others see 「権限がないため非表示」). Records are kept 2 years, then deleted automatically. A word search reads back through older records in steps: when the records read so far hold no match it says 「ここまでの記録には見つからない。「もっと読む」でさらに古い記録を探す」, and 「もっと読む」 searches further back; the total count appears only once the search has reached the oldest record. The filtered range can be exported as CSV / JSONL (「エクスポート」. Export, 「改ざんを検証」 and 「全件を検証」 are admins only; 「改ざんを検証」 checks only records added since the last check that passed, 「全件を検証」 checks every remaining record). JSONL has verification steps on its first line, so tampering can be detected wherever it is taken. 「アップロード前の検査」 is on the 「見つかったもの」 tab (one switch, 「アップロード前に個人情報・秘密情報を確かめる」; off by default; admins and managers switch it). While on, before a chat attachment or an upload from 「ナレッジを追加」, the browser itself checks the file for email addresses, phone numbers, マイナンバー, addresses, card numbers, API keys and .env secrets, and — next to a label such as 「氏名」「生年月日」「口座番号」「住所」「郵便番号」, or in the column under such a heading (a pasted Excel / CSV table included) — names, birth dates, bank accounts and addresses split into columns. No model is used and the file's content is not sent for the check. A hit opens 「個人情報・秘密情報が含まれている可能性があります」 with the place and a partly masked excerpt; the person chooses to send, stop, or 「マスクして送る」 (it never blocks). 「マスクして送る」 (text, CSV, Word, Excel, PowerPoint; not PDF) makes a copy in the browser with each hit replaced by its category name, checks it again, and sends the copy only if nothing remains. Images, zips and files that could not be read show 「未検査」 and go through; very large files show 「途中まで確認」. Sending with a hit records the file name and counts (never the values) in the audit log; sending a masked copy records nothing. 「情報保護の拡張」 is on the same tab: a Chrome / Edge extension that checks attachments and pastes on generative-AI sites (ChatGPT, Gemini, Claude, Copilot, Perplexity, NotebookLM, DeepSeek, Grok, Genspark) before they are sent, in the browser, with the same detectors and the same choices (send, stop, 「マスクして送る」); it works without connecting. The section is collapsed by default (press 「開く」 next to its heading; a link to #browser-extension opens it). Opened, it shows 「ストアを開く」, the extension ID, and the line to add to the Chrome policy ExtensionInstallForcelist ( |