Roster
Log inGet started

Privacy

Roster connects Claude to your Slack and Microsoft Teams, email (Outlook, Gmail, IMAP), calendars (Google, Outlook), Zoom recordings, Notion, Google Drive, Linear, GitHub, Jira, Confluence and HubSpot, read-only. This page lists exactly what we store and what we don't.

We don't store your content

Messages, emails, calendar events, meeting transcripts, files, pages, issues, CRM records and your search queries are fetched from the provider when Claude asks for them, passed straight back to Claude, and not saved by us. They never go into our database or our logs.

Email is read only from the inboxes you pick for the client you ask about. If an inbox is used for several clients, Roster only reads mail to or from that client's domains or addresses and drops anything else before it reaches Claude. Calendars and Teams chats follow the same rule, using the people invited to each meeting or in each chat. A Zoom account used for several clients only shows a client the meetings in that client's calendar invites. Linear teams, GitHub repos, Jira projects, Confluence spaces, HubSpot companies, Teams channels and Drive folders are read only for the one client you picked them for. Notion only shows Roster the pages you choose to share with it, and GitHub only the repos you install the Roster GitHub app on.

What we do store

  • Your account: email address and sign-in details, held by Supabase Auth.
  • Clients you create: name and slug.
  • Each connected workspace: your alias for it, the Slack workspace name and ID, your Slack user ID, the token's scopes, its status, and when we last reached Slack.
  • For every other connection (inbox, calendar, Teams, Zoom, Drive, Notion, Linear, GitHub, Atlassian site, HubSpot account): your alias for it, the account's email address, username or workspace/site name and ID, and its status.
  • The Linear teams, GitHub repos, Jira projects, Confluence spaces, HubSpot companies, Teams channels and Drive folders you picked for each client: their IDs and names.
  • Each client's email domains, which decide which mail belongs to it.
  • Each connection's access token (and refresh token, where the provider issues one), encrypted with AES-256-GCM. It is never shown in the dashboard or sent to Claude.
  • Each workspace's channel sharing choice: all channels, or the channel IDs you picked.
  • AI apps you've allowed, held by Supabase Auth so you can revoke them from the dashboard.
  • If you use a legacy connector URL, it is stored only as a one-way hash. We show the full URL once, when you create it, and can't show it again.
  • A request log for each tool call: which tool, which client, whether it worked, an error code if not, and how long it took. No queries, message text or results.

Who can see your Slack data

Only AI apps you sign in to and allow can read the workspaces connected to your account, and only the channels you share. Revoke an app on the dashboard and it loses access immediately. If you use a legacy connector URL, keep it private: anyone who has it can read the same way. Regenerate it on the dashboard if it leaks.

The Slack token acts as you, so it only sees channels you can already see. Answers in Claude are covered by your agreement with Anthropic.

Deleting your data

  • Disconnecting a workspace deletes its row and its encrypted token immediately.
  • Deleting a client deletes all of its workspaces and tokens.
  • Deleting your account (Settings in the dashboard) deletes everything above, including your sign-in.
  • You can also revoke access from Slack at any time by removing the app from the workspace.

Where it runs

The app is hosted on Vercel. The database and sign-in run on Supabase in the Tokyo region (ap-northeast-1). Slack data is fetched from Slack's API on demand.

Contact

Questions or deletion requests: prathamshirbhate1909@gmail.com.