AgentMail raises $6M to build an email service for AI agents

Just a couple of years ago, AI agents were mostly chatbots that could use basic tools. People were curious, but given concerns around reliability and security, as well as cost, the tech remained in the realm of early adopters.

How things have changed. Coding agents like Claude Code, Codex, and Cursor initially saw the most traction, spurring adoption among programmers around the world, but today we have people using AI agents to do everything from debugging at scale and building marketing campaigns, to managing calendars and scheduling meetings. OpenClaw‘s blockbuster debut earlier this year only sped things up, opening up access to AI agents by letting users run their own localized and personalized agent round the clock.

And if the tech industry is to be believed, AI agents are set to become as numerous as real people on the internet, using software and services, talking and shopping on your behalf, and generally automating a wide swathe of work.

AgentMail, a startup out of San Francisco, sees that future playing out for certain, which is why it has built an email service designed specifically for AI agents. The company provides an API platform that lets you give AI agents their own email inboxes, with support for two-way conversations, parsing, threading, labeling, searching, and replying.

The company on Tuesday said it had raised $6 million in a seed funding round led by General Catalyst, with participation from Y Combinator, Phosphor Capital, and angel investors Paul Graham, Dharmesh Shah (CTO of HubSpot), Paul Copplestone (CEO of Supabase), and Karim Atiyeh (CTO of Ramp).

Alongside the funding, AgentMail also announced an onboarding API that you can point your AI agent to so it can directly sign up and create an email inbox for itself. The platform also lets you set up and manage inboxes, permissions, allowlists, and API keys manually.

According to co-founder and CEO Haakam Aujla (pictured above, far right), AgentMail was built from the ground up to provide AI agents a similar inbox experience as people get with services like Gmail or Outlook — except without the UI elements humans need. (Note: The platform provides a perfectly human-usable interface, too, for managing the various agent inboxes, and reading and sending emails.)

“When you open Gmail, you have a bunch of threads, and inside each thread, you can have many messages; those messages can have attachments. You want to be able to label them, search them, filter them, reply, forward,” Aujla told TechCrunch. “We thought we wanted our agents to be able to do that, but they shouldn’t have to, you know, click buttons on a screen, because that’s pretty clunky for agents to do. They should just be able to make API calls.”

Since launching as part of Y Combinator’s Summer 2025 batch, the company has attracted tens of thousands of human users, and hundreds of thousands of “agent users,” Aujla said, as well as more than 500 B2B customers.

The initial days were slow, however, as AI agents hadn’t really taken off yet. AgentMail, therefore, focused on B2B use cases for companies that wanted to things like scale their email communications. But when OpenClaw (then known as Clawdbot) burst onto the scene in late January, AgentMail saw its user count triple that week, and quadruple in February as people started looking for a way to give agents an email inbox so they could do more on their own.

The timing was just right, as traditional email providers like Gmail impose rate and volume limits on their email APIs. AgentMail, meanwhile, provides a pretty generous free tier, in addition to paid plans and enterprise subscriptions.

But there’s an obvious issue with giving email inboxes to AI agents: it makes misuse easy. To counteract abuse, Aujla said AgentMail has a few systems: Agent inboxes can only send 10 emails a day unless they’re authenticated by a person; the platform imposes rate limits if it detects unusual levels of high activity from inboxes; monitors for bounce rates; and randomly samples new accounts to filter for sensitive keywords.

Aujla says beyond providing a way for bots to send and receive emails, AgentMail’s larger purpose is to serve as an identity layer for AI agents: “We want to give agents the ability to use email in the same way that humans do, right? But the idea is, what humans use email for is not even communication. It’s your identity […] There are several startups that are trying to build new identity protocols for agents, but our thesis is, let’s just use what already works for humans, and what already is so deeply integrated into the entire internet.”

“You give an agent an email address, [and] it can now use essentially any software service that already exists.”