Documentation

Getting Started

Salmex I/O is a personal AI operations platform that works for you. It connects to the AI models you choose, searches the web, remembers across conversations, and reaches you on every channel you use.

Quick start — Desktop

Download and install

Grab the latest .dmg from salmex.io. Open it, drag Salmex I/O to your Applications folder, and double-click to launch. macOS will ask you to confirm — click Open.

Set up the database

Salmex I/O needs a database to store your conversations and memories. The easiest option is Postgres.app — see the Database Setup guide for full instructions.

Configure an AI provider

The onboarding wizard walks you through setup. At minimum, add one AI provider — Anthropic (Claude), OpenAI (ChatGPT), Google (Gemini), or Ollama for fully local AI.

Start chatting

Open the Chat view and send your first message. Your agent can search the web, manage files, run tasks, and more. You can also connect your favourite channels — Telegram, WhatsApp, Discord, iMessage — so your agent is reachable wherever you are.


Basic configuration

AI providers

Salmex I/O works with multiple AI providers. Add one or more and choose your default — you can switch any time.

ProviderWhat you needNotes
AnthropicAPI keyClaude models. Recommended default.
OpenAIAPI keyGPT-4o, o1, o3, and newer models.
GoogleAPI keyGemini models.
OllamaNothing (runs locally)Fully local AI. No API key needed, no data leaves your machine.

Search engines

Give your agent the ability to search the web. Add at least one for research tasks.

EngineWhat you needNotes
PerplexityAPI keyAI-powered search with citations. Best for research.
BraveAPI keyPrivacy-focused web search. Good general-purpose option.
GoogleAPI key + Search Engine IDGoogle Custom Search.

Memory

Memory is automatic. Your agent learns your preferences, decisions, and important facts as you chat — and remembers them across every conversation, every channel, and every AI model. Switch from Claude to GPT to Gemini and back; mention something on Telegram and it's remembered everywhere. It's all the same agent, the same brain.


What's next

  • Setup & Installation — detailed database and onboarding guide
  • Features — chat, channels, memory, schedule, and settings
  • Plans — compare Free, Pro, CPU, and GPU options
  • Architecture — deep dive into how Salmex I/O works
  • Remote Access — deploy behind nginx, Caddy, or Cloudflare
  • Discord — get help, share what you've built, request features