AgentStack
Browse Sign in
Browse Why AgentStack Sell Docs
Sign in
MCP verified MIT Self-run

Agent Mail

mcp-wan2ted-work-agent-mail · by wan2ted-work

Disposable, programmable email inboxes on the fly — for humans and AI agents. Self-hostable, MCP-enabled. Powers email for AI agents at NotPeople.ai.

No reviews yet
0 installs
30 views
0.0% view→install

Install

$ agentstack add mcp-wan2ted-work-agent-mail

✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access Used
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets Used
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

View the full security report →

Verified badge

Passed review? Show it. Paste this badge into your README, it links to the public security report.

AgentStack Verified badge Links to your public security report.
[![AgentStack Verified](https://agentstack.voostack.com/badges/verified.svg)](https://agentstack.voostack.com/security/report/mcp-wan2ted-work-agent-mail)

Reliability & compatibility

Security review passed
0 installs to date
no reviews yet
3mo ago

Declared compatibility

Claude CodeClaude DesktopCursorWindsurf

Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.

Preview Execution monitoring

We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.

How agent discovery & health will work →
Are you the author of Agent Mail? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

📬 Agent Mail

Create disposable, programmable email inboxes on the fly — for humans and AI agents.

[](LICENSE) [](mcp) [](backend) [](frontend) [](CONTRIBUTING.md)

[](https://github.com/wan2ted-work/agent-mail/stargazers) [](https://github.com/wan2ted-work/agent-mail/commits/main) [](https://github.com/wan2ted-work/agent-mail/issues)


Agent Mail turns any domain you control into an endless supply of inboxes. Send mail to anything@.yourdomain.com and it instantly appears in the space tied to that key — no per-address setup, no mailbox provisioning. Point a whole custom domain at it and every address on that domain lands in one place.

It ships with a web UI for humans and an MCP server so AI agents can provision their own inboxes and read mail programmatically (e.g. to receive sign-up confirmations, OTPs, or notifications during automated workflows).

> Used in production. Agent Mail is the email infrastructure behind > NotPeople.ai — it gives fleets of AI agents disposable, > centrally-managed mailboxes for the sign-ups, confirmations, and OTPs they encounter > while operating. It's open-sourced here so anyone can run the same for their own agents.

Screenshots

Inbox — every address on a key flows into one searchable space (here: Reddit OTPs, verification, and login links an agent triggered).

Read a message — HTML or plain-text view, copy content, full headers.

Instance dashboard — manage keys, attach MX-verified custom domains, grab the API secret.

Why

  • On-the-fly addressesreport@acme.x.com, signup@acme.x.com, anything@acme.x.com

all route to the same space. Invent addresses as you go.

  • Agent-native — the MCP server lets a Claude/LLM agent create a mailbox and poll it for

the verification email it just triggered.

  • Self-hostable — bring your own domain and mail server, own all the data. MIT licensed.
  • Or fully managed — don't want to run a mail server? Use the hosted tier on mypm.cloud

and skip straight to the addresses.

How it works (60-second version)

Agent Mail does not speak SMTP/IMAP itself. A standard mail server (Postfix) receives mail and drops it into a Maildir; Agent Mail's worker polls that directory, parses each message, and routes it to the right space.

                                          ┌──────────────────────────────┐
  sender ──► MX of yourdomain.com ──► Postfix ──► Maildir/new/*.eml      │
                                          └───────────────┬──────────────┘
                                                          │ poll (every N s)
                                                          ▼
                                            ┌──────────────────────────┐
                                            │  Agent Mail backend       │
                                            │  • parse recipient        │
                                            │  • route by key / domain  │
                                            │  • store in PostgreSQL    │
                                            └─────────┬─────────┬───────┘
                                                      │ REST     │ REST
                                              ┌───────▼──┐   ┌───▼────────┐
                                              │ Web UI    │   │ MCP server │
                                              │ (Vue 3)   │   │ (agents)   │
                                              └───────────┘   └────────────┘

Routing. The recipient address decides the destination:

| Address pattern | Routes by | Example | |-------------------------------|------------------------------------|---------------------------------| | x@.yourdomain.com | the key (a subdomain label) | otp@acme.example.com | | x@customdomain.com | a verified custom domain | otp@acme.com |

Each "space" is an instance, identified by a secret UUID. Add keys/domains to an instance, and matching mail flows in. Mail that arrives before a key exists is stored as orphaned and auto-linked the moment you register the matching key.

> See [docs/architecture.md](docs/architecture.md) for the full data model and request flow.

Repository layout

| Path | What it is | |---------------|------------------------------------------------------------------------| | backend/ | aiohttp + PostgreSQL API and the Maildir polling worker (Python 3.10) | | frontend/ | Vue 3 + Vite + Pinia + Tailwind web client | | mcp/ | Model Context Protocol server exposing mailbox tools to AI agents | | docs/ | Architecture, setup, security, API and MCP guides | | docker-compose.yml | One-command local stack (db + backend + frontend [+ mcp]) |

Quick start (local)

git clone https://github.com//agent-mail.git
cd agent-mail
cp .env.example .env            # set POSTGRES_PASSWORD, EMAIL_DOMAIN, ...
docker compose up --build       # db + backend + frontend
  • Web UI:
  • API: (GET /health to check)

No mail server yet? Drop a raw .eml file into maildir/new/ and it'll be ingested within POLL_INTERVAL seconds. To receive real mail, point a domain's MX at your Postfix and have it deliver to the same Maildir — full walkthrough in [docs/getting-started.md](docs/getting-started.md).

Using it with an AI agent (MCP)

The MCP server lets an agent provision and read mail without touching the UI:

// Claude Code / Claude Desktop MCP config
{
  "mcpServers": {
    "agent-mail": {
      "command": "python",
      "args": ["-m", "agent_mail_mcp"],
      "env": {
        "AGENT_MAIL_API_URL": "http://localhost:8080",
        "AGENT_MAIL_EMAIL_DOMAIN": "example.com"
      }
    }
  }
}

Then an agent can call create_mailbox, get back signup@.example.com, trigger a flow that mails that address, and list_emails / get_email to read the result. See [docs/mcp.md](docs/mcp.md).

For AI agents

  • [AGENTS.md](AGENTS.md) — how an agent provisions a mailbox and reads mail; conventions.
  • [llms.txt](llms.txt) — machine-readable index of the project for LLMs/tools.
  • [examples/](examples/) — drop-in MCP config, a Python OTP flow, and curl recipes.

Want to try it instantly with sample data? make demo brings up the stack and seeds a mailbox with example emails.

Documentation

  • [Architecture](docs/architecture.md) — components, data model, routing internals
  • [Getting started / self-hosting](docs/getting-started.md) — DNS, Postfix, deploy
  • [Security](docs/security.md) — threat model and hardening checklist
  • [API reference](docs/api.md) — every endpoint, with request/response shapes
  • [MCP server](docs/mcp.md) — tools, configuration, agent recipes

Deployment models

| | Managed (mypm.cloud) | Self-hosted (this repo) | |---|---|---| | Run a mail server? | No | Yes (Postfix or similar) | | Get an address | x@.mypm.cloud instantly | x@.yourdomain.com after MX setup | | Custom domain | Point MX at mail.mypm.cloud, verify | Point MX at your own server | | Data ownership | Hosted | Entirely yours |

The managed tier exists so people can use Agent Mail without deploying anything. Self-hosting gives you full control — your DNS, your mail server, your database. See [docs/getting-started.md](docs/getting-started.md#two-deployment-models) for the trade-offs.

About

Agent Mail is built and maintained by the team behind NotPeople.ai, where it powers email for fleets of AI agents — disposable, centrally-managed inboxes for the sign-ups, confirmations, and one-time codes agents run into in the wild. We open-sourced it because every team building autonomous agents eventually needs the same thing: a programmable inbox an agent can create on the fly and read on its own.

If you're building AI agents, browser automations, or test harnesses that need to receive real email, this is for you. Star the repo, try the [MCP server](docs/mcp.md), and tell us what you build.

Contributing

PRs welcome — see [CONTRIBUTING.md](CONTRIBUTING.md). Security issues: see [SECURITY.md](SECURITY.md).

License

[MIT](LICENSE) © 2026 wan2ted

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet, be the first.

Versions

  • v0.1.0 Imported from the upstream source.