Product Concept: Nexus Mail – Self-Hostable Superhuman Alternative

1. Executive Summary

Premium email clients like Superhuman deliver speed, keyboard-first workflows, and AI-assisted triage—but they lock intelligence behind a subscription SaaS, route mail metadata and message content through vendor clouds, and leave teams with no control over model choice, data residency, or cost.

Nexus Mail is a self-hostable Superhuman alternative. It focuses on the same high-velocity email experience—split inboxes, shortcuts, snippets, reminders, and AI-powered reading/writing—while running on infrastructure you control. AI is not a locked vendor model: users add their own LLM API key (OpenAI, Anthropic, OpenRouter, Azure OpenAI, local gateways, or any OpenAI-compatible endpoint). Nexus Mail orchestrates prompts and tools; the key and the bill stay with the user.

2. The Challenge: Paid AI Email, Closed Stacks

Superhuman-class products solve real pain: inboxes are noisy, replies take too long, and “smart” features feel indispensable once adopted. The trade-offs are structural:

• SaaS lock-in: Workflow, AI quality, and pricing are controlled by one vendor. Features can change or disappear overnight.
• Data path opacity: Message bodies, threads, and sometimes contacts pass through third-party AI and indexing services—hard to justify for legal, finance, healthcare, or regulated startups.
• One model, one price: You cannot pick a cheaper model for triage, a stronger one for drafts, or a private endpoint for sensitive threads.
• No self-host path: Teams that want Superhuman UX on their VPC or home server have few open alternatives that still feel modern.

Nexus Mail targets people who want Superhuman speed and AI, not Superhuman as a landlord.

3. The Solution: Nexus Mail

Nexus Mail is a self-hosted email client for IMAP/SMTP (and modern providers via OAuth where available). Deploy it once—Docker, bare metal, or a small VPS—and connect mailboxes the same way you would any desktop client.

What makes it a Superhuman alternative:

• Keyboard-first UX: Command palette, Gmail-style shortcuts, blazingly fast navigation between threads and split views.
• Intelligent triage: Labels, priorities, and “Focus / Other” style split inboxes driven by rules plus optional LLM classification.
• AI that you own: Summarize threads, draft replies, rewrite tone, extract action items—only when a user-configured LLM key is present.
• Snippets, templates, and follow-up reminders for high-volume operators and founders.
• Self-hosted by default: Data, attachments cache, and search index live on your instance—not on a closed email SaaS.

Bring Your Own LLM Key

AI is an optional acceleration layer, not a forced cloud dependency.

1. Open Settings → AI / LLM.
2. Paste an API key and base URL (or pick a preset provider).
3. Choose models per job: e.g. a small model for classification and a larger one for long drafts.
4. Set limits: max tokens, which folders may use AI, and whether bodies leave the instance only as prompt context to your chosen endpoint.

No key means a full-featured mail client still works—search, shortcuts, snippets, offline cache—without AI. With a key, Superhuman-style intelligence turns on without a second AI subscription owned by the email vendor.

4. Product Experience

A. Speed Layer

• Instant open of recent threads from a local cache.
• Prefetch next messages while reading.
• Zero-mouse triage: archive, snooze, label, and reply without leaving the keyboard.

B. Split Inbox & Rules

• Priority vs everything else, VIP senders, and custom filters.
• Optional LLM assist: “Is this urgent?” / “Is this newsletter?” when the user enables AI for that mailbox.

C. AI Reading & Writing

• One-key thread summary and “what do they need from me?”
• Draft replies in your voice (tone presets + snippet injection).
• Rewrite: shorter, more formal, or more direct—using the user’s model and key.

D. Operator Tools

• Snippets with variables.
• Send later / follow-up bump.
• Shared team snippets and labels when multi-user self-host is enabled.

5. Engineering & Architecture

Nexus Mail separates mail transport, local intelligence, and optional remote LLM calls.

graph TD
    A[IMAP / SMTP / OAuth Providers] --> B[Nexus Mail Server]
    B --> C[Encrypted Store & Search Index]
    B --> D[Web / Desktop Client]
    D --> E[Keyboard-First UI]
    D --> F[AI Orchestrator]
    F --> G[User-Supplied LLM API Key]
    G --> H[OpenAI / Anthropic / OpenRouter / Local Gateway]
    C --> D

A. Self-Hosted Core

• Single-node or reverse-proxied deploy (HTTPS terminated by Caddy/Nginx/Traefik).
• Per-user mail credentials and encryption at rest for cached bodies where configured.
• Background sync workers for IDLE/push-friendly providers.

B. AI Orchestrator (BYO Key)

• Keys stored encrypted per user; never shared across tenants in multi-user mode.
• Prompt templates for triage, summary, and draft; no hard-coded vendor model ID.
• OpenAI-compatible client so local stacks (Ollama, vLLM, LiteLLM) work the same as cloud APIs.
• Explicit consent: AI features call out when message content will be sent to the configured endpoint.

C. Privacy Defaults

• No telemetric “improve the model with your mail” path.
• AI can be disabled globally by the instance admin.
• Optional allowlists: only certain labels or folders may hit the LLM.

6. Who It’s For

• Founders and operators who like Superhuman speed but refuse another opaque AI bill.
• Privacy-conscious teams that must keep mail on their own servers.
• Power users who already pay OpenAI/Anthropic and want those keys reused in email—not a third AI tax.
• Homelab and indie-hackers who self-host everything else and want email UX to match.

7. Impact & Positioning

Bottom line: Nexus Mail is Superhuman-class workflow without Superhuman as the middleman. Host it yourself; plug in your LLM key; keep speed, triage, and drafts under your control.

No resources added for this note.