AgentMail secures $6 million to provide AI agents with individual email inboxes.

      The San Francisco startup, supported by General Catalyst and Y Combinator, believes that email, rather than a new identity protocol, will be the means by which AI agents gain a foothold on the internet.

      Currently, AI agents can schedule your meetings, negotiate contracts, and manage support queues, but up until now, they have lacked a proper email address to operate from. AgentMail, a startup from Y Combinator’s Summer 2025 cohort, has secured $6 million in seed funding to address this gap. The funding round was led by General Catalyst, with contributions from Y Combinator and Phosphor Capital, along with angel investors such as Paul Graham, Dharmesh Shah (CTO of HubSpot), Paul Copplestone (CEO of Supabase), Karim Atiyeh (CTO of Ramp), and Taro Fukuyama.

      The company offers an API platform that provides AI agents with fully operational email inboxes, featuring real addresses that support two-way communication, threading, labeling, searching, replying, and parsing structured data from incoming messages. An inbox can be created with a single API call, requiring no OAuth flows, manual setup, or human intervention.

      The platform comes ready to integrate with LangChain, LlamaIndex, and CrewAI, and is compatible with any framework capable of making an API call. Additionally, AgentMail is releasing an onboarding API that allows an AI agent to register itself automatically by navigating to the platform, creating an inbox, and utilizing it without needing developer assistance. Remarkably, some agents are already taking advantage of this feature.

      “We’ve encountered an unexpected occurrence,” the company noted in its launch announcement. “Autonomous agents have begun signing up for AgentMail independently, discovering us through web searches, navigating to our site, and creating their own inboxes without developer involvement.”

      Co-founder and CEO Haakam Aujla, a former quantitative researcher at Optiver, started the company with Michael Kim, who has a background in autonomous vehicles at Nvidia, and Adi Singh, who previously held investment roles at Accel, StepStone Group, and Flex Capital. Their theory is that the identity issue for AI agents does not necessitate a completely new protocol.

      Email already serves as the identity layer of the internet, deeply integrated into every existing service and application. “If you provide an agent with an email address,” Aujla told TechCrunch, “it can utilize essentially any software service that already exists.”

      The timing of the product has been influenced by both internal and external events. AgentMail was launched in August 2025 and spent its initial months focusing on business clients needing to scale their email communications, with gradual progress.

      However, in late January 2026, OpenClaw gained widespread attention. This platform allowed users to operate their own local AI agents continuously, leading to an immediate and significant demand for the type of agent infrastructure that AgentMail was developing. User numbers tripled the week after OpenClaw's rise and quadrupled the following month. Currently, the company reports tens of thousands of human users, hundreds of thousands of agent users, and over 500 B2B customers, according to Aujla.

      The variety of use cases has expanded alongside the increase in users. Supply chain teams are employing agents to coordinate carriers and resolve freight exceptions via email in real time. Loan collection agents manage payment reminders and follow-ups, while procurement bots engage in negotiations with vendors. The common challenge faced is scale: traditional email providers like Gmail were designed for individual use and impose rate limits and pricing per inbox that hinder large-scale agent deployments.

      A valid concern arises from providing AI agents with their own email addresses, as this could facilitate misuse. Aujla openly addressed this issue and outlined several safety measures: agent inboxes are restricted to ten outbound emails per day unless verified by a human; the platform sets rate limits for unusual activity patterns; bounce rates are monitored; and new accounts are screened for sensitive keywords. Whether these controls will suffice as the platform expands is a question the company will face as its user base of agents grows.

      For Yuri Sagalov of General Catalyst, the investment is based on a straightforward insight regarding how agent identity will function in practice. “Email is central to identity on the internet,” Sagalov stated. “Conventional identity services were not designed with agentic use cases in mind, and AgentMail is addressing that part of the stack, commencing with email.”

      Aujla presents the company’s goal in similar terms but projects further ahead. Email is just the beginning, not the end. As agents increasingly take on tasks currently performed by humans, they will require not only inboxes but also credentials, reputation, and trust—the comprehensive framework of an online identity. AgentMail’s strategy is to build outwards from this most universal component of that framework. In this view, the next billion internet users are already on their way; they just need a place to receive their email.