Launching today
OpenMail
Give your AI agents an identity through email
98 followers
Give your AI agents an identity through email
98 followers
Your agents sign up for tools, receive verification codes and reply to customers, mostly from your Gmail or a shared sending pool where one bad actor gets everyone flagged. OpenMail gives every AI agent its own inbox on your domain with its own sending reputation, for one API call each. Incoming mail reaches your code over webhook or WebSocket, attachments arrive as text, and there's a hosted MCP server and a Hermes plugin. Built and hosted in the EU, free for 3 inboxes.
Free Options
Launch Team / Built With
BRCKS ChatThe chat app for building sites. Nothing gets lost.
Promoted
Hi Product Hunt, Jokūbas here, co-founder of OpenMail with @efemeralizacija .
We started in March because every agent we saw was using a borrowed email address. It signed up for tools and answered customers from someone's Gmail, or from a shared sending domain where one bad tenant got everyone's mail flagged. That falls apart once you run more than a few agents, and faster when the agents belong to your customers.
OpenMail gives each agent its own inbox: a real address on your domain, inbound delivered to your code over webhook or WebSocket, attachments turned into text your model can read, and a sending reputation that belongs to that inbox alone, so one agent misbehaving doesn't take the others down with it. There's a hosted MCP server and a Hermes Agent plugin, and all of it runs in the EU.
Most of this year went into receiving and keeping tenants apart: isolation and abuse containment, the part that never makes it into a demo video and that you need by your second customer.
Launch-day offer: make an account today and enter Launch2026 in billing before Friday 6 November for three months of Pro. The free tier is 3 inboxes and 3,000 emails a month either way.
If you've built something that needs email, tell me what broke with your current setup; that is what we build next.
And ask the hard questions about deliverability, abuse and pricing. Armandas and I are here all day and will answer every one, with numbers where we have them.
Docs: docs.openmail.sh. Thanks for having a look.
@jokubas_vaiciulis_ @efemeralizacija This could become an important part of the AI agent ecosystem. How do you handle identity verification for agents ?
@jokubas_vaiciulis_ @dipanshu_kushwaha5 Hmm, the question is double sided, so I'll try to answer both. Simply put, the agents that are using OpenMail use our email addresses to log in but use other tools like browsers to get in. As for verification to whom the agent belongs to, we have the person's email as a touchpoint but it's not visible to others, but maybe it should - good idea though!
Love it, where do you think differentiation will come from in this space? What’s next?
@veeti_vallin there are many ways we can expand further, most of the requests that we get are about making calendar access to agents, human - agent collaboration in email client, password manager for agents, etc! If you have any additional thoughts, let us know.
I hate when I set a prompt for an AI agent to code all night but it gets stuck on some email verification. Tons of time wasted. Could this help me out?
@domantas_vanagas Yes, that’s pretty much the exact case we built it for. You set up an OpenMail account once and connect it to your agent (the MCP server works with Claude, Cursor and Codex). From then on the agent can make its own inbox, sign up with that address and read the verification email itself, so the run doesn’t stall at 3am waiting for you. It won’t get past a CAPTCHA or an SMS code, but the email step it handles on its own. The free tier is 3 inboxes, which is plenty for an overnight run.
This looks like a very interesting addition to the agent ecosystem. I was curious how you handle spam / marketing emails. Do agents also alert users about marketing emails?
@atul_ranjan2 We have a classifier that looks into the type of emails are sent and spam, marketing emails usually arrive to our "Spam" folder which does not inform the agents to get a reaction on. You can still look into them manually or ask your agent for a look up.
How does it compare to agentmail?
@edas_lakavicius AgentMail is better in many ways, but our core focus is on our users' reputation management and infrastructure hosted in EU. You can throw anything at us, and we'll handle it appropriately - isolate users per their bounce/complaint rates, making sure other inboxes don't hurt the others if you run AI agent platform, like Polsia.
Inbound retries seem worth making very visible when an agent is on the other end. I'd want it to recognize the same message arriving twice before it sends two replies.
@oleksii_sekundant Agreed, with an agent on the other end that’s the case that bites. Every retry of the same incoming email carries the same event ID, so your agent can tell it’s seen it before. And if a retry still gets through, the send call can be told “this reply is for that event”, so OpenMail won’t send a second one.