Lumbox (formerly AgentMailr)
by Lumbox
Real email inboxes and login credentials for AI agents, behind one API
Lumbox, launched as AgentMailr, is email and identity infrastructure for AI agents. One API call creates a real inbox, and every incoming message is parsed into JSON with one-time passcodes and verification links extracted as fields. It bundles an encrypted credential vault and a stealth browser so agents can clear signups and logins unattended.
Lumbox is hosted email and identity infrastructure built for autonomous software rather than for people. It launched as AgentMailr and was renamed in March 2026 after a trademark notice from the Y Combinator-backed AgentMail, which raised $6M that same month; the vendor states the product, API and existing inboxes were unchanged by the rename, and agentmailr.com now redirects to lumbox.co. A single API call provisions a real, agent-owned mailbox, and every message that arrives is parsed server-side into structured JSON with one-time passcodes, magic links, verification links and backup codes extracted as first-class fields, so an agent never regex-scrapes a message body. A long-poll endpoint blocks until the code lands, removing the polling loop teams otherwise hand-write around SES or IMAP. Around that inbox the product assembles the rest of the signup-and-login path: an AES-256-GCM encrypted credential vault that injects passwords without exposing plaintext to the model, a Steel Browser integration with anti-bot handling and CAPTCHA solving for form filling, and boundary markers wrapped around email content so models do not mistake message text for instructions. Access is via REST, a JavaScript/TypeScript SDK, IMAP/SMTP, webhooks, or an MCP server that drops straight into Claude Code and Cursor. Custom sending domains are supported with MX, SPF and DKIM verification. It is a solo-founder product by engineer Deepanshu Kumar, running its API and mail servers on Hetzner infrastructure in Germany, and it competes head-on with AgentMail and indirectly with Amazon SES, Postmark and Resend for teams willing to assemble the layers themselves.
The engineering team building autonomous agents that must register for, verify and log into third-party services without a human in the loop — the point at which a shared team mailbox stops working.
Every agent gets its own real inbox whose verification codes arrive as parsed JSON on a blocking endpoint, so the signup-and-OTP step stops being custom mail plumbing.
At a Glance
- Category
- Agent Development
- Pricing
- Freemium, Subscription, Usage-based
- Target Market
- Enterprise Developers, AI Engineers, Startup Founders, Automation Engineers
- Deployment
- Cloud-only, API-based
- Headquarters
- Germany
- Team Size
- 1-10
Key Features
- ✓On-demand agent inboxes
Create a persistent or throwaway mailbox with one API call so each agent owns a unique address and replies route back to that specific agent.
- ✓Automatic OTP and verification-link extraction
Incoming mail is parsed server-side into JSON fields for one-time codes, magic links, backup codes and summaries, removing brittle regex parsing from agent code.
- ✓Long-poll wait-for-code endpoint
A blocking endpoint returns the moment a verification code arrives, so agents skip hand-written polling loops and arbitrary sleep timers entirely.
- ✓Encrypted credential vault
Passwords are stored under AES-256-GCM and injected at use time, so plaintext secrets never enter the model's context window or logs.
- ✓Steel Browser automation
A stealth Chromium with anti-bot protections and CAPTCHA solving lets agents complete the web signup and login flows that the inbox unlocks.
- ✓MCP server, SDK and IMAP/SMTP access
The same inboxes are reachable over REST, a JavaScript/TypeScript SDK, IMAP/SMTP, webhooks, or an MCP server wired into Claude Code and Cursor.
- ✓Custom sending domains
MX, SPF and DKIM verification means outbound agent mail leaves the customer's own domain rather than shared vendor sending infrastructure.
- ✓Prompt-injection boundary markers
Email content is wrapped in explicit delimiters so a model reading a message does not treat attacker-supplied text as its own instructions.
Capabilities
Use Cases
- •Autonomous SaaS account signup
An agent registers for a trial, receives the confirmation email, extracts the verification link and finishes onboarding without any human involvement.
- •Unattended MFA and login
Agents log into third-party dashboards by pulling one-time passcodes from their own inbox instead of borrowing an employee's personal mailbox.
- •Per-agent attribution and audit trails
Giving every agent a distinct address keeps replies, bounces and audit records attributable rather than collapsing into one undifferentiated team inbox.
- •Outbound agent notification and reply handling
Agents send status updates or prospecting mail from a verified customer domain, with threading and reply routing handled entirely by the API.
- •Research behind login walls
Scraping agents clear registration and CAPTCHA gates using the bundled browser and credential vault, reaching sources anonymous crawlers cannot open.
Ideal For
Best For
- ✓Giving each AI agent a dedicated, persistent email address instead of sharing a human or team mailbox
- ✓Automating account signup and email verification flows end to end
- ✓Retrieving one-time passcodes and magic links programmatically for unattended logins
- ✓Wiring email actions into Claude Code or Cursor through an MCP server
- ✓Sending outbound agent notifications or replies from a customer-verified domain with DKIM/SPF/DMARC
Not Ideal For
- ✗Regulated enterprises that require SOC 2, ISO 27001 or HIPAA attestations before procurement — Lumbox publishes no compliance certifications or trust page
- ✗Teams already operating Amazon SES, Postmark or their own IMAP parsing, where Hacker News commenters argued the existing human-email stack already covers the job
- ✗High-volume human-facing marketing email, where dedicated deliverability platforms like SendGrid and Postmark have far deeper tooling and reputation management
- ✗Buyers who need vendor stability signals — this is a single-founder product that has already changed names once under legal pressure
Integrations
Deployment
Market Analysis
Pros
- ✓One API call provisions a real mailbox and parsed OTP/link fields arrive as JSON, removing the most tedious part of agent email plumbing
- ✓The blocking wait-for-code endpoint eliminates hand-written polling loops and timing guesswork
- ✓Transparent, self-serve pricing with a genuinely usable free tier (3 inboxes, 500 received emails/month)
- ✓MCP server means it drops into Claude Code and Cursor without custom glue
- ✓Custom domains with DKIM/SPF/DMARC keep outbound mail on the customer's own sending reputation
Cons
- ✗No independent review coverage exists at all — G2, Capterra, TrustRadius and Product Hunt carry no listing, and the Show HN drew 7 points and three comments
- ✗Hacker News commenters questioned the premise outright: 'Why does the agent need an API different than the ones we have for humans? (SMTP, IMAP/POP, AWS SES, etc..)'
- ✗The same thread raised deliverability and domain-reputation risk when many agents send from shared infrastructure; bring-your-own-domain is the mitigation but pushes DNS setup back onto the customer
- ✗No published SOC 2, ISO 27001 or GDPR documentation, which will stall enterprise security review
- ✗Single-founder project that has already rebranded once under a trademark notice, and a commenter reported the footer GitHub link was broken
Pricing
Free
$0
- ✓3 agent inboxes
- ✓500 emails received/month, 100 sent
- ✓OTP and verification-link extraction
- ✓MCP server and REST API access
Starter
From $9/mo
- ✓10 agent inboxes
- ✓5,000 received / 2,000 sent per month
- ✓Webhooks
- ✓Custom domains with MX, SPF and DKIM
Pro
From $29/mo
- ✓50 agent inboxes
- ✓25,000 received / 10,000 sent per month
- ✓AI email categorization (bring your own key)
- ✓Thread routing, contact lists and marketing emails
Scale
From $99/mo
- ✓250 agent inboxes
- ✓100,000 received / 50,000 sent per month
- ✓Priority support
- ✓SLA guarantee and dedicated deliverability
Enterprise
Contact for pricing
- ✓Dedicated infrastructure
- ✓Custom SLAs
- ✓Direct contact via contact@lumbox.co
List pricing is published in full and self-serve, which is unusual in agent infrastructure: $0 for 3 inboxes and 500 received emails a month, then $9 Starter, $29 Pro and $99 Scale at 250 inboxes and 100,000 received emails, with a custom Enterprise tier for dedicated infrastructure and SLAs. Metering is by inbox count plus received and sent email volume rather than by seat, so cost tracks agent fleet size. AI email categorization on Pro is bring-your-own-key, meaning model spend sits outside the subscription.
Security & Compliance
Sources
This page was written from 6 sources, 2 on domains other than lumbox.co.
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
Kitesurf
An agent-first browser that runs in V8 isolates on Cloudflare Workers
Natural
Payments infrastructure that lets AI agents hold, send and collect money
AgentPrizm
Governed, persistent memory and reusable skills for AI agents — over REST API and MCP
Bespoke Labs
Reinforcement-learning environments that train reliable, production-ready AI agents