An e-mail based AI co-worker
Hi IH,
after a long break I want to share my idea: an AI agent you talk to via email - just forward anything to [email protected] and it remembers, reasons, and replies.
If you want to test it, send an email from https://uff.email by "request your agent" and add to message that you are from IH community.
Happy to discuss the idea — feedback very welcome.
Trending on Indie Hackers
The interesting opportunity isn't giving AI an email address—it's turning email into a persistent interface for long-term collaboration with an agent. I'd keep validating whether people return because they can access AI from anywhere or because email becomes the place where the agent's memory naturally compounds over time.
Exactly right. The 'AI with an email address' is the hook, but the product is 'email as a persistent interface where agent memory compounds.'
That's the interesting distinction.
The shift from "AI access through email" to "email becoming a persistent interface for an agent" changes the product category entirely.
I had one thought about how that memory layer could affect adoption and retention, but I'd rather discuss it in the context of what you're building than reduce it to another thread comment.
What's the best email to reach you on?
Any address that you find on uff.email website will be ok
Thanks! I’ve just sent it over.
Looking forward to hearing your thoughts whenever you have a chance.
the email-as-interface choice is the strong part, and i'd lean harder into WHY it's strong. email is async and reply-native, so it's a natural approval surface: the agent proposes "here's the draft, here's what i'd do" and you just reply "send it" from your phone. that human-in-the-loop step is exactly what people are nervous about with autonomous agents, and email hands it to you for free. i'd make that the pitch, not the "it remembers everything you forward" part, since forwarding-to-remember is a chore and storing everything tends to bury the signal more than surface it.
Great comments, both - really appreciate the engagement. It also shows I didn't nail the value prop I had in mind, so thanks for pushing on it.
The intent isn't an agent that acts for me (approving and sending drafts) nor an agent that just stores everything I forward. It's a private collaborator that only talks to me - no inbox, no replies to anyone else.
Threads are session contexts where we iterate together until I'm ready to act.
It reads, analyzes, reminds, helps me form a better response (using build in skills) - but I do the sending, I do the acting. Email works because it's async, private, and already where I live - and over time the agent builds a compounding context of my work without ever crossing into acting on my behalf.
Really appreciated both your comments - they made me rethink the narrative from scratch. Updated the landing page (uff.email) with a few key shifts:
From "AI with an email address" to "private AI collaborator" - the agent works with you, not for you. No inbox, no replies to anyone but you.
From "forward anything" to "ask or forward" - you can just ask a question, not just forward docs.
From "it remembers, reasons, replies" to "reads, analyzes, remembers — so you save time and respond better" - the human stays in control, the agent amplifies.
Added a pricing comparison - $25–50/hr for a VA vs $9/mo for the agent. Same async workflow, always on.
Cut Skills & Built-with sections - too much noise, not enough signal.
Every thread is a private session - iterate until you're ready, then you act.
Still early, but the reframe clicked instantly with the existing users.
Thanks again for the comments and feedback - it genuinely sharpened the pitch.