Sign-in and security
Two-step sign-in, your sessions, AI tools signed in as you, where your data lives
Settings › You › Sign-in is yours alone. It shows how you sign in, every browser you are signed in on, and every AI tool that acts as you.
Two-step sign-in. After your password, or Google or Microsoft, MindBacklog asks for a six-digit code from an authenticator app on your phone (Google Authenticator, 1Password, Authy or any app that makes six-digit codes).
- To turn it on: Turn on (if you didn't sign in during the last 15 minutes, enter your password first, or sign in again), scan the QR code, then enter the code the app shows.
- You then get 8 recovery codes, shown once. Copy them or download them as a .txt file and keep them away from your phone. Each one signs you in once if you lose the phone. MindBacklog keeps only a fingerprint of each, so nobody, support included, can show them to you again. Make new codes (with a code from your app) replaces the whole set.
- The code is asked for on each new sign-in, Google and Microsoft included (MindBacklog can't see their own two-step), and when you accept an invitation to another workspace.
- After five wrong codes in a minute (or twenty in an hour) you wait before trying again.
- Turning it off needs a code from your app or a recovery code, and signs your other browsers out.
- Lost the phone and every recovery code? Write to support. They check it's you before they reset it, and the reset is in the workspace's audit log.
Require two-step for everyone (admins, Settings › Security). Only once your own two-step is on. The page says who has it on. Someone without it is sent to set it up when they next open the workspace. It holds for AI tools too: they can't sign a tool in to the workspace, and a tool they signed in before waits until they turn two-step on.
Your active sessions. Every browser signed in as you: "Chrome on Mac · this device · now", "Safari on iPhone · 2 hours ago". Sign out signs one browser out on its next click, even if it had "keep me signed in". Sign out everywhere else signs out all of them but this one. A password reset signs every browser out. The location shows as "location not known" for now.
AI tools signed in as you. Claude Code, Cursor, claude.ai and other tools that support MCP sign-in can sign in as you, with no token to copy:
- Claude Code: in your project,
claude mcp add --transport http mindbacklog <your MindBacklog address>/mcp, then type/mcp, pick mindbacklog and Authenticate. - Cursor: add the address (no token) to
~/.cursor/mcp.json; Cursor shows Needs login: click it. - The browser opens MindBacklog: "{Tool} wants to use MindBacklog as you". Choose the workspace and one product or all you open, then Approve. The tool can read what you can read, and change what you can change (a status, the product map, a build report), checked on every call. It never deletes anything.
- A tool stays signed in while you use it (it renews itself; after 30 days without use it asks again). Sign it out in Settings › Your AI tools or You › Sign-in: its next call is refused.
- Tools that can't sign in still use a personal token (Settings › Your AI tools), as before. Your tokens keep working.
- If a tool asks you to sign in again, MindBacklog shows a page first: nothing signs you out until you press Sign out and sign in again.
For admins (Settings › Security and Team).
- AI tools signed in to {workspace}: every person's signed-in tools and tokens, by name (never a token), with Sign out on each.
- Let AI tools update status (off by default): an editor's signed-in tool, on a product they can only read, may tick off stories and mark features shipped, never anything else. A viewer's AI tools only read. People who can edit always can.
- Team: each member reads Can edit, Can read and use AI tools or Can read, with Turn AI tools on/off. It's free: no seat, no price. Turning a member's AI tools off stops their tools and tokens on the next call; turning them on again brings them back. Viewers who held a token when the new MindBacklog turned on keep their AI tools.
Where your data lives (Settings › Security). The region is stated only once it has been confirmed; until then the page says "Not confirmed yet". Mind's AI models are run by the providers listed on the page, where they run them. I need the EU records that your workspace needs an EU region, and why and by when. There is no EU region today.
Asking for AI tools. When someone whose AI tools are off tries to make a token, the workspace's admins are told once a day: on the bell (with who asked) and as an item in their digest. Turn their AI tools on in Settings › Team.