MCP-native · email, SMS & voice for AI agents

Give your AI agent its own inbox, phone number, and voice line.

Inboxr is the MCP server that lets an agent spin up a real mailbox, receive the verification email, and pull the OTP or magic link — on its own, mid-task. One config block in Claude, Cursor, or Claude Code and your agent can sign up, confirm, and log in like a human. REST, SMTP, and webhooks are there too when you want them.

Works in Claude Desktop · Cursor · Claude Code
One config block — then just ask

Add the MCP server. Your agent does the rest.

Paste one block into Claude, Cursor, or Claude Code and the agent can create inboxes, wait for mail, and pull OTPs on its own. Prefer to wire it yourself? The same surface is one SDK import, one Python call, or one curl away.

// 1. Add the MCP server to your agent (claude_desktop_config.json,
//    .cursor/mcp.json, or .claude/settings.json):
{
  "mcpServers": {
    "inboxr": {
      "command": "npx",
      "args": ["-y", "@jassra25/inboxr-mcp"],
      "env": { "INBOXR_API_KEY": "inb_live_..." }
    }
  }
}

// 2. Now just ask the agent — it drives the tools itself:
//
//   "Sign up for the app at staging.example.com with a fresh
//    inbox, click the verification link, and tell me the
//    dashboard URL you land on."
//
// Under the hood the agent calls:
//   create_inbox()        -> a real mailbox, instantly
//   wait_for_message()    -> blocks for the verification email
//   extract_link()        -> pulls the magic link, opens it
//
// No glue code. The agent owns the whole signup-and-confirm loop.
How an agent gets its own channel

The same loop, whatever the channel.

An agent asks for a channel, gets a real one provisioned on the spot, waits for the inbound message, and extracts what it needs — all through MCP tool calls. Flip between channels to see the flow.

Your AI agent
Claude · Cursor · Claude Code
provisions
Live email channel
narwhal-7f3@getinboxr.app
1
create_inbox()

Agent gets a real, live mailbox — no human picks an address.

2
wait_for_message()

Long-polls until the verification mail lands, ~200 ms after MX.

3
extract_link()

Pulls the magic link straight out of the HTML body.

Signed up, confirmed, and logged in — no human in the loop.
Live in production

MCP-first. Then REST, SMTP, and webhooks when you want them.

Real email, SMS, and voice — running today on getinboxr.app — with an MCP server built so an AI agent can drive a whole signup-and-confirm flow on its own. REST, SMTP, and webhook surfaces are right there for the code you write by hand.

MCP for AI agents

Drop @jassra25/inboxr-mcp into Claude Desktop, Cursor, or Claude Code. Tools for email, SMS, and voice — create_inbox, wait_for_message, extract_otp, extract_link, sms_*, voice_* — let an agent run signup-and-confirm flows end-to-end.

Wire up an agent

REST API

Bearer-authenticated /v1 endpoints for inboxes, messages, sending, and waiting. Try every call live in the in-portal playground.

Open the playground

SMTP submission

smtp.getinboxr.app:587 with STARTTLS. Issue a per-inbox credential from /v1/inboxes/{id}/smtp-credentials and drop it into nodemailer / lettre / msmtp.

See SMTP docs

Webhooks

Push every inbound message to your CI runner within ~200 ms of arrival. HMAC-SHA256 signed, 5-attempt retry with exponential backoff.

Configure webhooks
For AI agents
# Claude Desktop / Cursor / Claude Code
{
  "mcpServers": {
    "inboxr": {
      "command": "npx",
      "args": ["-y", "@jassra25/inboxr-mcp"],
      "env": {
        "INBOXR_API_KEY": "inb_live_...",
        "INBOXR_API_URL": "https://api.getinboxr.app"
      }
    }
  }
}

Restart your agent and it can create_inbox, wait_for_message, extract_otp, and extract_link without writing a line of glue.

Built for agents first

Everything an agent needs to act like a real user.

Agent-owned channels

An agent calls create_inbox / sms_create_inbox and gets a real, live mailbox or number it controls — provisioned mid-task, no human in the loop.

wait_for_message, not polling glue

One MCP call blocks until the mail or SMS lands, then extract_otp / extract_link hands back exactly what the agent needs. No regex plumbing to write.

DKIM / SPF / DMARC inspection

Every message exposes auth-results + the full RFC822 source. Confirm you're emulating production correctly.

Open-core mail engine

Run the entire stack yourself — Postfix + Dovecot + OpenDKIM — with the same API surface. Pay us only when you want the cloud layer.

Per-project rate limits + usage

Hit a quota? Clear dashboard, predictable overage pricing. No surprise invoices.

OIDC + SAML SSO

Ship enterprise from day one. Google / Microsoft / Okta / custom SAML — plus SCIM for Scale + Enterprise.

Pricing

Start free. Pay when you scale.

Every plan ships with the full API, webhooks, and the web UI. Higher tiers add SSO, retention, and compliance. Need Enterprise? Get in touch.

Free
Enough to wire up a test suite.
$0forever
  • 100 messages / month
  • 5 active inboxes
  • 1k API calls / month
  • 7-day retention
  • Community support
Most popular
Dev
One engineer. Everything you need.
$19per month
  • 10,000 messages / mo
  • 50 active inboxes
  • 50k API calls / mo
  • Webhooks with retries
  • 30-day retention
  • Google / GitHub SSO
  • Email support
Team
For teams that ship every day.
$49per month
  • 100k messages / mo
  • 500 inboxes
  • 500k API calls / mo
  • Unlimited team members
  • 90-day retention
  • Microsoft + Okta OIDC
  • Tenant branding
  • Priority support
Scale
High-volume + compliance.
$199per month
  • 1M messages / mo
  • Unlimited inboxes
  • 5M API calls / mo
  • 1-year retention + legal hold
  • SAML + SCIM
  • Custom domain (inbox.yourdomain.com)
  • Audit export
  • SLA + dedicated support

Overage: $0.001 per message, $0.0001 per API call. Self-hosting free forever — run the engine yourself.

Give your agent an inbox of its own.

Real email, SMS, and voice your agent can provision and read on its own — through one MCP server. Free to start, no credit card.