In just the past few years, AI agents have evolved far beyond simple chatbots with basic tool access. While initial interest was tempered by worries over reliability, security, and expense—keeping the technology in the hands of early adopters—the landscape has transformed dramatically. Early adoption was driven by coding assistants like Claude Code, Codex, and Cursor, which gained popularity among developers globally. Today, however, people employ AI agents for a vast range of tasks, including large-scale debugging, creating marketing campaigns, and managing calendars and meeting schedules.

The breakthrough debut of OpenClaw earlier this year accelerated this trend significantly by enabling users to operate their own personalized, localized agents continuously. Industry predictions now suggest AI agents will soon be as ubiquitous as human users online, autonomously utilizing software, conducting conversations, making purchases, and automating a broad spectrum of work.

Convinced of this impending future, the San Francisco-based startup AgentMail has developed an email service tailored specifically for AI agents. The company offers an API platform that allows users to equip AI agents with dedicated email inboxes, complete with functionality for two-way communication, email parsing, conversation threading, labeling, search, and automated replies.

EMBED_PLACEHOLDER_0

On Tuesday, the company disclosed a $6 million seed funding round. The investment was 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).

Concurrent with the funding news, AgentMail introduced an onboarding API. This allows developers to direct their AI agents to the service to autonomously sign up and create an email inbox. The platform also supports manual setup and management of inboxes, permissions, allowlists, and API keys.

According to co-founder and CEO Haakam Aujla, AgentMail was engineered from the start to give AI agents an inbox experience comparable to human services like Gmail or Outlook, but without the graphical user interface elements that people require. (The platform does include a fully functional interface for humans to manage agent inboxes and handle emails.)

“When you open Gmail, you see multiple threads, and within each thread, there can be numerous messages containing attachments,” Aujla explained. “We wanted our agents to have that same capability, but without the need to click buttons on a screen, which is inefficient for them. They should simply be able to execute API calls.”

Since its launch as part of Y Combinator’s Summer 2025 cohort, the company has attracted tens of thousands of human users and hundreds of thousands of “agent users,” along with over 500 B2B customers, Aujla noted. Growth was initially slow, as AI agents had not yet reached mainstream adoption. Consequently, AgentMail first concentrated on B2B applications for businesses seeking to scale their email operations.

However, when OpenClaw (originally named Clawdbot) emerged in late January, AgentMail’s user base tripled that same week and quadrupled in February. This surge occurred as individuals sought ways to provide email inboxes to their agents, enabling greater autonomy. The timing proved fortuitous, as traditional email providers like Gmail enforce strict rate and volume limits on their APIs. In contrast, AgentMail offers a quite generous free tier, supplemented by paid plans and enterprise subscriptions.

A significant concern with providing email inboxes to AI agents is the potential for misuse. To combat abuse, Aujla stated that AgentMail has implemented several safeguards: agent inboxes are limited to sending 10 emails per day unless a human authenticates them; the platform applies rate limits upon detecting unusually high activity; it monitors bounce rates; and it randomly samples new accounts to screen for sensitive keywords.

Beyond merely enabling bots to send and receive email, Aujla describes AgentMail’s broader mission as establishing an identity layer for AI agents. “We aim to allow agents to use email just as humans do,” he said. “But the concept is that for humans, email’s primary function isn't just communication—it’s your identity. Several startups are attempting to build new identity protocols for agents, but our approach is to leverage what already works for humans and is deeply embedded across the entire internet.”

“Once you give an agent an email address,” Aujla concluded, “it can then access virtually any existing software service.”