The privacy-first AI agent
with its own inbox and phone number.
Delegate conversations, not your accounts. Plum never connects to your email or reads your history - you loop it in one conversation at a time, the way you'd brief a human assistant. And it's live in one click.
Free to start. No inbox access. No app. Texting arrives soon.
It starts knowing nothing. That's the point.
A human assistant doesn't get your email password on day one - they get looped in. Forward this. Handle that. Here's the context. Plum works exactly the same way.
Its own address
rosie@agent-plum.com
Your assistant is a colleague you email - not software wired into your accounts. There is no "connect your inbox" step, because it doesn't need one.
You choose every share
Forward a thread. Cc it in. Or just tell it.
What you send, it knows. What you don't, it never sees - your history stays yours.
Memory you can open
Plain files, on your dashboard.
It remembers what you've shared - and you can read, edit, or delete any of it, any time.
Most AI email tools begin with "connect your entire inbox." Plum begins with a name.
Live in one click
- Name it. Your assistant gets its own address - like rosie@agent-plum.com.
- Click the link we email you. That click brings it to life - no password, no setup, nothing to connect.
- Loop it in. Email it tasks, forward what matters, cc it into threads. It only ever knows what you share.
What it does
Ask anything
"Plan a three-course dinner for six on Saturday."
Replies in about a minute, like a sharp assistant on email.
Your calendar
"What does my Friday look like?"
Connect Google Calendar and choose per calendar: read-only, read & write, or nothing.
On a schedule
"Send me a briefing every morning at 8."
It sets up its own recurring tasks from a plain sentence.
In your threads
Cc it into any conversation.
It stays quiet until you address it, and briefs you privately when something needs you.
It remembers
Tell it once.
Preferences, people, and context persist - every conversation makes it better.
It learns skills
Repeat tasks become playbooks.
Your agent writes its own reusable skills as it works - inspectable and editable by you.
Private by design
Need-to-know isn't a setting you toggle - it's how Plum is built. A few more walls that come standard:
Only people you approve can talk to it
Cold email from strangers is dropped before the AI ever runs. You keep the contact list.
It can only message people you approve
Structurally - the tools refuse other recipients. Not a policy, a wall.
Strangers in your threads get watch-only mode
Notes and a private heads-up to you. Never replies, never actions.
Hard daily spend cap
Every agent pauses at its budget and tells you - a runaway can't run away.