Documentation

Getting Started

Download to first conversation in under 5 minutes. Salmex I/O is a personal AI operations platform that runs on your Mac. It connects to the AI models you choose, searches the web, remembers everything across conversations and channels, and answers to no one but you.

Everything below is free

The entire platform works at $0 for personal use. No trial, no credit card, no time limit. You can run Salmex I/O with your own API keys and your own database — forever. Pro and credit packs are optional upgrades for users who want managed inference, multi-channel, and commercial use.

1. Install

Download the app

Grab the latest .dmg from salmex.io. Requires macOS 15 (Sequoia) or later on Apple Silicon.

Drag to Applications and launch

Open the DMG, drag Salmex I/O to Applications, double-click to launch. macOS will ask you to confirm — click Open.

Gatekeeper

If macOS blocks the app, go to System Settings > Privacy & Security and click Open Anyway. This only needs to happen once.

2. Set up your database

Salmex I/O stores your conversations, memories, and tasks in PostgreSQL. The onboarding wizard gives you two options:

Bring your own database

Already have PostgreSQL running? Click I have my own database and enter your connection details. Works with local installs (Postgres.app, Docker, Homebrew) or cloud databases (Neon, Supabase).

Requires PostgreSQL 17+ with pgvector and uuid-ossp extensions.

Most users should choose built-in storage. It starts automatically with the app, handles migrations, and stores everything locally. If you want full control or already run PostgreSQL, see the detailed database setup guide.

3. Choose your AI

The onboarding wizard asks how you want to run AI inference. This is separate from your plan — every option below works on the Free tier.

Bring your own key

Free — pay your provider directly

  • Use Anthropic (Claude), OpenAI (GPT-4o, o3), or Google (Gemini)
  • Paste your API key — the app verifies it instantly
  • Switch providers any time without losing memory
  • Add multiple providers and choose per conversation

Click Add key during onboarding. You'll need an API key from your provider.

Ollama (100% local)

Free forever — runs on your Mac

  • Run open-source models locally — zero data leaves your machine
  • No API key, no account, no internet required
  • Works with any model Ollama supports (Llama, Mistral, Gemma, etc.)

Install Ollama first, then select it during onboarding.

You can mix and match

These aren't mutually exclusive. Use managed inference for most tasks, add your own Anthropic key for Claude, and run Ollama locally for privacy-sensitive work — all in the same app, all sharing the same memory. Change any time in Settings > AI.

Give your agent the ability to research the web.

Inference choiceSearch setup
Managed inferenceBrave Search included — nothing to configure. You can add your own Perplexity or Google key for more.
BYOK / OllamaAdd your own search API key. Supported: Perplexity (AI-powered, best for research), Brave (privacy-focused), Google (Custom Search). You can skip and add later.

5. Start chatting

That's it. You're in.

Open the Chat view and send your first message. Your agent can search the web, read and write files, run shell commands, manage tasks, and remember everything you tell it — across every conversation, every channel, and every AI model.

A few things to try:

  • Ask it to remember something — "Remember that I prefer TypeScript over JavaScript" — then start a new conversation and ask what it knows about you.
  • Give it a research task — "Find the latest pricing for AWS Lambda in eu-west-2 and summarise it."
  • Schedule something — "Every weekday at 9am, check Hacker News for AI news and send me a summary."
  • Connect a channel — Add Telegram so your agent reaches you on your phone. Same memory, same brain.

Trial & pricing

The platform and the inference are priced separately. The platform (Free or Pro) controls features like channels, remote access, and commercial use. Inference (managed or BYOK) controls how your AI runs.

FreePro — $9/mo
Memory, safetyUnlimitedUnlimited
Schedules3 activeUnlimited
ChannelsDesktop + 1Unlimited simultaneous
Remote accessDIY (bring your own tunnel)Managed, zero-config
Commercial usePersonal onlyPermitted
BYOK inferenceUnlimitedUnlimited
Managed inference14-day free trial, then credit packs from $5 (100 queries included with Pro)

When your trial or credit pack runs out, the app falls back to your own API keys — you're never locked out. See Plans for the full comparison.


What's next