Skip to content

Profiles & access

Three separate questions live under this page, and it helps to keep them separate: where your hub can be reached from, who can sign in to manage it, and what each individual connected tool is allowed to do once it’s in.

Set during the first-run setup, and changeable later in the dashboard under Settings › Access & sign-in:

  • Just this device or network. Nothing outside your own network can reach it, full stop. This is the safe default, and the right choice if every AI tool you use also runs on your own machine or network.
  • Reachable from the internet, sign-in required. For tools that connect from a company’s own service rather than from your device — a phone app, a web-based assistant — which need an address they can actually reach. Sign-in turns on the moment this does; there is no way to expose your memory to the internet without it.
  • Reachable from the internet, dashboard included. The same as above, plus the management dashboard itself is reachable too, not just your memory’s connection point. A stricter checklist runs before this is allowed, and it’s the right choice for fewer situations than the option above — most people who want to reach their memory from the internet don’t also need the dashboard exposed.

Whichever you pick, Connect your AI tells you, per tool, exactly what’s needed for that specific tool to reach you — some need nothing more than “just this device,” others need the internet-reachable option turned on.

The dashboard itself — the screen where you browse your memory, approve proposals, add tools, and change any of this — has its own sign-in, separate from anything an AI tool uses to connect. One person is set up as the administrator during first-run setup, right in the browser. palaia keeps exactly one such account: to change its password later, run palaia oauth set-password where the hub runs, or connect a sign-in provider (GitHub, or your company’s OpenID Connect) instead — then that provider’s allow-list decides who may sign in.

Every AI tool you connect gets its own named connection — “Claude Code on my laptop,” “the office assistant” — rather than one shared password everyone uses. Each one can be limited to exactly what it needs:

  • Which memory or memories it can see, if you keep more than one.
  • Whether it can only look things up, or also save and change things.

This matters most when a tool sits somewhere less trusted than your own laptop — a shared service, something you’re trying out — where you’d rather it could look things up but never write anything. Revoking one connection takes effect immediately and doesn’t touch any of the others.

A tool’s connection status is checked independently by that tool and by your hub, and the two checks don’t always agree at the exact moment you sign in — a tool can briefly report “not connected” right before it actually connects successfully. Troubleshooting & FAQ covers this and the other rough edges worth knowing about up front.