Show HN: MailPilot – Freedom to go anywhere while your agents work
3 days ago
What is this? A local TUI (like Claude, Codex, Gemini, OpenCode or Copilot) that wraps your agent and sends each turn to an email in a nice format.
What does it enable? It lets you email your agents. And them email you. So They Keep Working on the tasks you want, and You go and do What You Want. No need to stay at your desk. Be free.
I built this for me, but thought others would find it useful, so I turned it into a product. I want get outside, and away from my desk, but still have the agents work.
The accidental killer feature: You can CC your team. If you forward the thread to a coworker, their reply goes straight to the agent context too. It turns a local session into an async multiplayer thread.
Works out of the box with Claude, Codex, Gemini, Copilot, and OpenCode. Happy to answer questions!
> Local and private
> Your agent runs on your machine. We only relay the messages.
How can this be private if this intermediate service is sending and receiving all the emails back and forth
Just in that your agent runs on your local machine, has access to your local filesystem, and no code execution happens on our cloud, and that we don't look at or store the emails. Pure relay, so it’s just as private as business collaboration on regular email in that sense.
It's a paid product, you are not the product. We have 0 interest in your email content or data. Only in making it easy for you to run your agents without being stuck on your console.
Are the emails end to end encrypted (PGP or S/MIME where you/your server don't have the keys) or just in transit (TLS)? That would make the difference between "we can't look at your emails" and "we choose not to look at your emails".
9 replies →
I do a very similar thing but clunkier with Tailscale, ssh, zellij. Can ssh in from my phone too with Termux. But email sounds better - then I can actually get notified when it needs me!
I have to say, $12/mo feels steep. It's a minor improvement over what I have now. Compared to other $10+/mo services, this one feels pretty light.
I have claude code hooks that send local computer notifications when action is required or processing finished. And when I step away from computer, I get those notifications through pushover. Then I login on phone to ssh (mosh) with Termius and connect to the tmux session running claude. I use this approach when watching TV with the family and laptop is not appreciated on the sofa. :)
Every time I read these "I've managed to control Claude Code from my phone posts", they come with some variation of "so that I can continue being on my computer" during some other activity. It's a very personal decision, but feels like on of these points where people should re-evaluate. Just because we can, doesn't mean we should.
> it was so good that I caught myself coding from my phone while out with friends… and decided that this is something I should stop, more for mental health than anything.
https://steipete.me/posts/2025/shipping-at-inference-speed
2 replies →
This is smart and the right way!
It is pretty light. Maybe the pricing is wrong, idk. Early days so just figuring things out. The email advantage is as you say -- I always found Terminal on phones very taxing on typing, etc. Email is so much lighter on cognitive load etc. Thanks for having a look and for your feedback!
Oh, this isn’t the popular Mac email client, Mail Pilot (https://www.mailpilot.app/)? I’m sure that won’t confuse anyone.
Everyone, please, google your proposed product names before actually buying a domain.
Another good idea is to check a trademark DB, because "Mail Pilot" is actually a registered trademark of Mindsense LLC.
This is a great setup. I was building agents in 2023 with email addresses, and it was very effective for onboarding tech laggards to generative AI. It also resulted in very low durable use rates. Users would onboard then stop using quickly. Christenson ftw again.
One thing that can be handled much better today than back then was MIME / attachment types and having agents assemble "fancy" email structures irt.
https://github.com/realityinspector/ATAT
I haven’t done attachments yet, but that could be useful.
I think to really make this work You really have to know what you’re doing with your tasks and your agents because you have to be able to manage them from a distance with sort of minimal context just the email updates, which are summaries so it’s very hands off…so you either need the project you’re working on to fit that or you have to be really deep in your project so that you can manage it from a distance. It just for me is the correct separation of concern where like the agent is handling the details and they’re hidden from you.
I don’t see this as a gentle introduction to AI for people who have no idea what they’re doing about it or about their task. I see this is more like the natural extension or progression for power users who don’t want to waste time with their terminal when they don’t need to.
Nice. Would be great to have this on Slack or Whatsapp
No affiliation and didn't get around to use it yet, but clawdbot[1] does WhatsApp, Telegram, Slack, Discord, Signal, iMessage, Microsoft Teams, and WebChat.
[1]: https://github.com/clawdbot/clawdbot
A much simpler setup can achieve "I want [coding agent messages] on slack"
1 reply →
Ya, I agree that's something. Maybe in future!
I saw this concept only 8 days on this very site ( https://news.ycombinator.com/item?id=46517458#46523962 )
Looking forward to it.
Missed that! Thanks for the link. Definitely feels like this kind of idea is zeitgeist rn.
Coincidental that the domain was registered on the same day... it's fine to say you were inspired to build from the discussion.
Why do your privacy and terms state they were updated on February 28, 2025?
1 reply →
Considering emails are as private as postcards, I don't know how I could rely on such service to prompt anything even remotely sensitive.
Just like regular email in that sense. We don't do any encryption beyond standard HTTPS and we don't look at the data. It's a paid service, you're not the product.
Really cool! Started something in the same area as well that is fully open source: www.runagentos.com
This is such low hanging fruit. Coding agents can be orchestrated. Controlled, rerun, and tested. Are we just coming up with ways to not be at our desks working? Why not scale out agents on a cluster to do more work? Why email instead of SMS with a web endpoint? I can access each agent of mine via a dashboard from anywhere in the world.
Not trying to poo poo, just saying all it takes is Claude Code to introduce this and you’re done.
There's a sense in which you're right that it could be cloned easily by a provider. But that's also an advantage in that we work acorss providers, they're not going to create a service that interacts with each other. So it's more of a fit for someone outside the AI providers to do it, but we'll see.
>> Why email instead of SMS with a web endpoint
Pretty sure this was vibe-coded in a few days based on a discussion that was on HN a few days back and few people mentioned it would be nice to code on the go via email.
Oh, sorry, your pretty sure is wrong then. I created it myself. Intuition, an idea came into my mind. No discussion, no thread. Just me, and whatever I was tuend into. Basically like all my ideas I turn into products. Too bad you didn't know that, now you do.
> I can access each agent of mine via a dashboard from anywhere in the world.
What do you use for this?
I use a central server behind a vpn and oauth that gives me access to my "farm". A product I wrote to manage fleets of docker containers that's now being used to manage fleets of docker containers running a tty wrapper I also wrote to send everything over. The backend to this is go+htmx+websockets. The tty wrapper runs OpenCode / Claude / Codex / etc when it starts up and is given access to a repo via volume mount. From there, the "agent" either gets instructions from the websocket or it pulls tasks from Linear. Think ECS but each service instance is an agent.
I can login, create a task, look at task results, approve PR's, deny PR's, retask an agent, or take one over for my own exploratory endeavors.
1 reply →
Also, I made a Product Hunt: https://www.producthunt.com/products/mailpilot