RMS邮件

3. Configure required variables inside the .env file:

3. Configure required variables inside the .env file:

Images are pulled from Docker Hub: maxramas/rms-mail:m-latest + m-ui-latest

```

Once started, open your browser and navigate to: 👉 http://localhost:3000

For HTTPS production (aaPanel, nginx, Caddy), see reverse-proxy.md — proxy :3000 only on Mono.

🚀 Quick Start

2. Run configuration script

./configure.sh

2. Set up your environment variables

cp .env-m.example .env

3. Configure your `ENCRYPTION_KEYS` or `ENCRYPTION_KEY` and `JWT_SECRET` inside the .env file

4. Copy the Mono-specific compose configuration

cp docker-compose-m.yml docker-compose.yml

2. Set up your environment variables

cp .env-u.example .env

4. Copy the Unified-specific compose configuration

cp docker-compose-u.yml docker-compose.yml

2. Set up your environment variables

cp .env-mp.example .env

4. Copy the Mono Pro-specific compose configuration

cp docker-compose-mp.yml docker-compose.yml

🧠 Unified Multi-Account Workflow

Unified edition solves one of the biggest missing features in self-hosted email: Real multi-account workflows. Features: unified inboxes, unified project groups, cross-account search, cross-account bulk actions, unified notifications, centralized AI workflows. Designed for: agencies, freelancers, operations teams, infrastructure engineers, support-heavy environments.

---

🔌 MCP Server & IDE Integration

RMS Mail ships with a native MCP server. Use your mailbox directly from Cursor, Zed, Claude Desktop, custom agents, orchestrators, and IDE-integrated workflows. Available capabilities: search_emails (fully isolated by MCP API Key context — keys scoped per mailbox account) get_email * email_agent (Natural language email operations with native tool-calling maps)

MCP works behind HTTPS reverse proxies (Mono and Unified): SSE and JSON-RPC on /mcp/* through the UI entrypoint.

This is not an "AI wrapper" integration. Your mailbox becomes part of your agent ecosystem.

💬 Deep Telegram Integration

RMS Mail includes a deeply integrated Telegram bot. Capabilities: inbox summaries instant notifications AI-assisted chat mailbox search quick actions (/archive, /delete, /reply) workflow automation * signed webhook payloads ({ "event": "email.received", "email": { ... } }) with HMAC-SHA256

The same AI + mailbox system works consistently across browser UI, Telegram, MCP, and agents.

---

⚡ Native Database FTS & Performance Pipeline

RMS Mail does not rely on slow IMAP search or memory-heavy external indexing sidecars. Every email passes through a zero-overhead pipeline ensuring instant access even inside massive directories.

┌─────────┐     ┌────────────────────────┐     ┌───────────────────────────────┐     ┌─────────┐
│  IMAP   │ ──▶ │   SQLite FTS5 (Mono)   │ ──▶ │  Go Memory Pre-parsing        │ ──▶ │   UI    │
│ Server  │     │ Postgres GIN (U, MP, T)│     │  zstd Raw Payload Compression │     │ (React) │
└─────────┘     └────────────────────────┘     └───────────────────────────────┘     └─────────┘

Pipeline:

1. Batch-based IMAP synchronization (500 UID chunks) with streaming MIME to disk.
2. Metadata normalization and cross-language UTF-8 strict sanitization.
3. Native full-text index generation (SQLite FTS5 virtual engine / PostgreSQL tsvector + GIN).
4. Keyset cursor pagination (is_pinned, date_sent, id) — O(1) depth at any inbox offset.
5. PostgreSQL production tuning (v3.1.0): covering index (folder_id, is_read, is_muted, is_pinned DESC, date_sent DESC, id DESC) eliminates sort on filtered reads; BRIN index on date_sent for compact time-series aggregations; connection pool min(20, CPU*4) prevents OOM in Docker; ANALYZE emails after bulk sync prevents Seq Scan; autovacuum_vacuum_insert_scale_factor = 0.05 triggers vacuum promptly on insert-heavy workloads.
6. Real-time UI virtualization (TanStack Virtual + measureElement).

Result: typically sub-100 ms text search, instant filter counting, fast bulk operations on six-figure mailboxes.

---