AgentMail

AgentMail gives each agent a full, persistent inbox — not just a way to fire off outbound mail. It supports two-way conversations with threading, attachments, drafts, scheduled send, labels, and full-text search across every inbox in an org, plus real-time events over webhooks and websockets so an agent can react to incoming mail instead of polling. It also exposes IMAP/SMTP for compatibility with existing mail tooling, and an MCP server so agent frameworks can pick up inbox actions as tools. TechCrunch reported the founders built this specifically because Gmail's API imposes rate and volume limits and requires OAuth, both of which are awkward for software that provisions inboxes programmatically at scale (https://techcrunch.com/2026/03/10/agentmail-raises-6m-to-build-an-email-service-for-ai-agents/). For a buyer choosing between 'bolt an agent onto Gmail' and 'give the agent its own mailbox,' this is a meaningfully different product shape, not a repackaged transactional-email API.

Published in full on the pricing page, verified 2026-08-28 (https://www.agentmail.to/pricing): Free — $0/month, no card required, 3 inboxes, 3,000 emails/month, 100 emails/day, 3 GB storage. Developer — $20/month, 10 inboxes, 10,000 emails/month, 1,000/day, 10 custom domains, 2 seats, email support. Startup — $200/month, 150 inboxes, 15,000 emails/day, 150 custom domains, 10 seats, and a SOC 2 report becomes available at this tier. Enterprise is custom (unlimited inboxes, white-label, EU-region cloud, BYO cloud, dedicated IPs, OIDC/SAML SSO) — no published floor. The jump from Developer to Startup is 10x the price for roughly 15x the inbox ceiling, which is the point at which a growing agent fleet needs to budget for a step change rather than a smooth scale-up. Early-stage startups can apply for a free month on the Startup tier via a Tally form linked from the pricing page.

Handing autonomous software its own email address invites spam and account-takeover risk that a human-operated inbox doesn't. According to the founders (via TechCrunch, https://techcrunch.com/2026/03/10/agentmail-raises-6m-to-build-an-email-service-for-ai-agents/), new agent inboxes default to a 10-outbound-emails-per-day cap unless a human authenticates the account, the platform applies rate limiting on unusual activity spikes, monitors bounce rates, and randomly samples new accounts for sensitive keywords. A WorkOS writeup of a company demo (May 2026) independently describes the same allow-list-plus-daily-cap design as the default when an agent self-provisions an inbox from a prompt (https://workos.com/blog/agentmail-mcp-night-4-email-identity-layer-agents). These are vendor-described mechanisms, not independently audited, but the design — restrictive defaults, loosened deliberately — is the right shape for the risk, and it's unusual for a company this young to have thought about it this explicitly.

AgentMail is a San Francisco company founded in 2025, went through Y Combinator's Summer 2025 batch, and lists a team of 10 as of 2026-08-27 (https://www.ycombinator.com/companies/agentmail). It raised a $6M seed round in March 2026 led by General Catalyst, with Y Combinator, Phosphor Capital, and angel investors including Paul Graham, HubSpot CTO Dharmesh Shah, Supabase CEO Paul Copplestone, and Ramp CTO Karim Atiyeh participating (https://techcrunch.com/2026/03/10/agentmail-raises-6m-to-build-an-email-service-for-ai-agents/). The founders told TechCrunch the product has 'tens of thousands' of human users, 'hundreds of thousands' of agent users, and more than 500 B2B customers — these are vendor figures with no independent corroboration and should be read as claims, not facts. Practically: this is a company six months past its first institutional raise, with a small team supporting infrastructure that agents rely on for identity and communication. That's not a disqualifier, but it's the right context for weighing uptime, support responsiveness, and longevity against a more established provider.

AgentMail publishes DKIM/SPF/DMARC support, signed webhooks, MFA, and social SSO across all tiers on its pricing page. A SOC 2 report is available starting at the $200/month Startup tier; OIDC/SAML SSO, EU-region cloud, and BYO-cloud deployment are Enterprise-only and require contacting sales. There is no public security or trust-center page — agentmail.to/security returns a 404, and no independent trust-center listing (e.g., a Vanta or Drata public page) was found. A status page exists and is live (https://status.agentmail.to, Better Stack-hosted). For a buyer with a compliance checklist, that means the SOC 2 report itself is gated behind a paid tier and a sales conversation rather than published — reasonable for a young company, but it means you can't independently verify compliance posture before paying.

The most substantive third-party writing found is a SaaStr blog post by Jason Lemkin describing how Claude recommended AgentMail to him, he signed up, and never evaluated a second option (https://www.saastr.com/i-need-agentic-email-claude-said-try-agentmail-so-i-did-and-never-looked-at-anything-else/). It's a genuine testimonial — 'I'm a paying customer' — but it's framed almost entirely around AI-answer-engine strategy, not a comparative evaluation of deliverability, reliability, or support quality against alternatives like Postmark, Resend, or SendGrid. Beyond that and the TechCrunch funding writeup, we found no independent benchmarks, uptime data, or comparative reviews. The company's GitHub org (37 repos) shows active daily commits as of late August 2026 across its Python, Node, CLI, MCP, and docs repos (https://github.com/agentmail-to), which is a reasonable proxy for active maintenance, but it is not a substitute for independent product evaluation.