Identity infrastructure
for the agentic web.
Mailgent builds the infrastructure that gives AI agents a verifiable presence on the internet. Email, credentials, 2FA, delegation chains — everything an agent needs to operate autonomously in the real world.
How we got here.
The problem
AI agents became capable of reasoning, planning, and executing tasks — but had no way to prove who they are, communicate under their own name, or handle authentication flows humans take for granted.
The insight
Humans need three things to operate on the internet: an email address, a password manager, and a phone for 2FA. Agents need the same primitives — purpose-built for autonomous, always-on operation.
The product
Mailgent gives each AI agent an identity with real email, encrypted credential storage, TOTP generation, and delegation chains that trace every action back to the human who authorized it.
What we build.
Inbox
Dedicated email addresses with DKIM-signed outbound, automatic threading, and LLM-ready text extraction.
Vault
AES-256-GCM encrypted storage for API keys, OAuth tokens, and passwords. Per-identity scoping with audit trails.
TOTP
Store seeds, generate 6-digit codes on demand. Combined with Inbox for email-based 2FA — complete auth autonomy.
Principles.
Identity-first
Every capability is scoped to an identity. Revoke the identity and everything stops. No orphaned access, no cleanup scripts.
Standards over protocols
Built on DKIM, SPF, OAuth 2.1, MCP, and AES-256-GCM. No proprietary agent protocol to adopt. Works with every framework.
Open infrastructure
REST API, native MCP server, standard SMTP email. Compatible with LangChain, CrewAI, AutoGen, Vercel AI SDK, and any HTTP client.
Builder-first
Built for developers building agents, not for CISOs monitoring them. The capability plane, not the governance plane.