Only a few years back, AI agents primarily functioned as chatbots with limited tool integration. While the public showed interest, issues regarding dependability, security, and expense kept this technology largely within the domain of innovators.
The landscape has dramatically shifted. Initially, coding agents such as Claude Code, Codex, and Cursor gained significant momentum, prompting widespread adoption among developers globally. Now, individuals leverage AI agents for diverse tasks, including large-scale debugging, crafting marketing strategies, and handling calendar and meeting arrangements. The highly successful launch of OpenClaw earlier this year further accelerated this trend, democratizing access to AI agents by enabling users to operate their own localized and customized agents continuously.
Should predictions from the tech industry prove true, AI agents will soon be as prevalent as human users online, interacting with software and services, conducting conversations, making purchases, and broadly automating numerous tasks on our behalf.
A San Francisco-based startup named AgentMail is confident in this future, leading them to develop an email service specifically tailored for AI agents. Their API platform enables the creation of dedicated email inboxes for AI agents, facilitating two-way communication, message parsing, conversation threading, labeling, searching, and replying.
On Tuesday, the company announced it successfully secured $6 million in seed funding. This round was spearheaded by General Catalyst, with additional contributions from Y Combinator, Phosphor Capital, and notable angel investors including Paul Graham, Dharmesh Shah (HubSpot’s CTO), Paul Copplestone (Supabase’s CEO), and Karim Atiyeh (Ramp’s CTO).
Concurrent with the funding announcement, AgentMail introduced an onboarding API, allowing AI agents to autonomously register and establish their own email inboxes. Furthermore, the platform offers manual capabilities for configuring and overseeing inboxes, permissions, allowlists, and API keys.
Haakam Aujla, co-founder and CEO (shown above, far right), stated that AgentMail was fundamentally designed to offer AI agents an email experience akin to human services like Gmail or Outlook, but without the user interface components typically required by humans. (It’s worth noting that the platform also includes a fully functional human interface for managing agent inboxes, as well as for composing and reading emails.)
Aujla explained to TechCrunch, “When you access Gmail, you encounter multiple threads, each potentially containing numerous messages and attachments. You expect to label, search, filter, reply to, and forward these messages. We aimed for our agents to perform these same functions, but without the cumbersome need to click on screen elements; instead, they should simply execute API calls.”
Since its debut during Y Combinator’s Summer 2025 cohort, the company has gained tens of thousands of human users, hundreds of thousands of ‘agent users,’ and over 500 B2B clients, according to Aujla.
Initially, growth was modest because AI agents had not yet achieved widespread adoption. Consequently, AgentMail initially concentrated on B2B applications, serving companies seeking to enhance their email communication scale. However, following the emergence of OpenClaw (formerly Clawdbot) in late January, AgentMail’s user base tripled within that week and quadrupled by February, as individuals increasingly sought methods to equip agents with email capabilities for greater autonomy.
This timing proved optimal, given that conventional email services like Gmail enforce rate and volume restrictions on their email APIs. In contrast, AgentMail offers a considerably generous free tier, alongside various paid plans and enterprise subscriptions.
However, a clear challenge arises from providing email inboxes to AI agents: the potential for misuse. To combat this, Aujla explained that AgentMail employs several safeguards: agent inboxes are limited to sending 10 emails daily unless authenticated by a human; the platform implements rate limits upon detecting abnormally high activity from any inbox; it monitors bounce rates; and new accounts are randomly sampled to screen for sensitive keywords.
Aujla asserts that AgentMail’s broader objective, beyond enabling bots to exchange emails, is to establish an identity layer for AI agents: “Our goal is to empower agents to utilize email similarly to humans, correct? Yet, the fundamental role of email for humans isn’t merely communication; it’s about identity. While numerous startups are developing novel identity protocols for agents, our approach is to leverage an existing solution that is effective for humans and deeply embedded across the internet.”
“By providing an agent with an email address, it gains the capability to interact with virtually any existing software service.”