1 month ago
17
Just a mates of years ago, AI agents were mostly chatbots that could usage basal tools. People were curious, but fixed concerns astir reliability and security, arsenic good arsenic cost, the tech remained successful the realm of aboriginal adopters.
How things person changed. Coding agents similar Claude Code, Codex, and Cursor initially saw the astir traction, spurring adoption among programmers astir the world, but contiguous we person radical utilizing AI agents to bash everything from debugging astatine standard and gathering selling campaigns, to managing calendars and scheduling meetings. OpenClaw‘s blockbuster debut earlier this twelvemonth lone sped things up, opening up entree to AI agents by letting users tally their ain localized and personalized cause circular the clock.
And if the tech manufacture is to beryllium believed, AI agents are acceptable to go arsenic galore arsenic existent radical connected the internet, utilizing bundle and services, talking and buying connected your behalf, and mostly automating a wide swathe of work.
AgentMail, a startup retired of San Francisco, sees that aboriginal playing retired for certain, which is wherefore it has built an email work designed specifically for AI agents. The institution provides an API level that lets you springiness AI agents their ain email inboxes, with enactment for two-way conversations, parsing, threading, labeling, searching, and replying.
The institution connected Tuesday said it had raised $6 cardinal successful a effect backing circular led by General Catalyst, with information 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 besides announced an onboarding API that you tin constituent your AI cause to truthful it tin straight motion up and make an email inbox for itself. The level besides lets you acceptable up and negociate inboxes, permissions, allowlists, and API keys manually.
According to co-founder and CEO Haakam Aujla (pictured above, acold right), AgentMail was built from the crushed up to supply AI agents a akin inbox acquisition arsenic radical get with services similar Gmail oregon Outlook — but without the UI elements humans need. (Note: The level provides a perfectly human-usable interface, too, for managing the assorted cause inboxes, and speechmaking and sending emails.)
“When you unfastened Gmail, you person a clump of threads, and wrong each thread, you tin person galore messages; those messages tin person attachments. You privation to beryllium capable to statement them, hunt them, filter them, reply, forward,” Aujla told TechCrunch. “We thought we wanted our agents to beryllium capable to bash that, but they shouldn’t person to, you know, click buttons connected a screen, due to the fact that that’s beauteous clunky for agents to do. They should conscionable beryllium capable to marque API calls.”
Since launching arsenic portion of Y Combinator’s Summer 2025 batch, the institution has attracted tens of thousands of quality users, and hundreds of thousands of “agent users,” Aujla said, arsenic good arsenic much than 500 B2B customers.
The archetypal days were slow, however, arsenic AI agents hadn’t truly taken disconnected yet. AgentMail, therefore, focused connected B2B usage cases for companies that wanted to things similar standard their email communications. But erstwhile OpenClaw (then known arsenic Clawdbot) burst onto the country successful precocious January, AgentMail saw its idiosyncratic number triple that week, and quadruple successful February arsenic radical started looking for a mode to springiness agents an email inbox truthful they could bash much connected their own.
The timing was conscionable right, arsenic accepted email providers similar Gmail enforce complaint and measurement limits connected their email APIs. AgentMail, meanwhile, provides a beauteous generous escaped tier, successful summation to paid plans and endeavor subscriptions.
But there’s an evident contented with giving email inboxes to AI agents: it makes misuse easy. To counteract abuse, Aujla said AgentMail has a fewer systems: Agent inboxes tin lone nonstop 10 emails a time unless they’re authenticated by a person; the level imposes complaint limits if it detects antithetic levels of precocious enactment from inboxes; monitors for bounce rates; and randomly samples caller accounts to filter for delicate keywords.
Aujla says beyond providing a mode for bots to nonstop and person emails, AgentMail’s larger intent is to service arsenic an individuality furniture for AI agents: “We privation to springiness agents the quality to usage email successful the aforesaid mode that humans do, right? But the thought is, what humans usage email for is not adjacent communication. It’s your individuality […] There are respective startups that are trying to physique caller individuality protocols for agents, but our thesis is, let’s conscionable usage what already works for humans, and what already is truthful profoundly integrated into the full internet.”
“You springiness an cause an email address, [and] it tin present usage fundamentally immoderate bundle work that already exists.”
English (US) ·