Personal AI agent
MayrosMayros

An AI agent that acts for you.

Lives on your devices and responds in WhatsApp, Telegram, Slack, and 20+ channels. Choose your AI —Claude, GPT, or local models—, run 50+ actions, and log every one. Self-hosted and open-source: your data never leaves you.

20+ Channels·20+ AI Models·50+ Actions
Mayros
The problem

Today's AI forgets, locks you in, and doesn't answer for its actions.

Assistants forget

Every conversation starts from scratch. You repeat your context over and over, and the AI never truly learns who you are.

They live in a single app

A chatbot stuck in a tab. It doesn't follow you to WhatsApp, Slack, or your terminal — it's never where you are.

You never know when they're making things up

They respond with confidence… and sometimes hallucinate. Without logs or proof, you can't trust what they did for you.

In action

Talk to it like a colleague. It gets the work done.

Ask in any channel and Mayros executes the action with your tools, confirms through your preferred channel, and logs what it did.

WhatsAppTelegramSlack
Schedule coffee with Ana on Thursday at 10:00 ☕
Done. I confirmed it via Slack.
action executed · calendar

your AI (Claude · GPT · local) · self-hosted · every action logged

Capabilities

An agent platform, not a chatbot.

Where you already chat
WhatsApp, Telegram, Slack, Discord, Teams, iMessage, Signal, and 20+ channels. No new app to install: your agent comes to you.
Choose your AI
Claude, GPT, Gemini, Ollama (free local models), Bedrock… 20+ providers. Your key, your control, your AI cost.
50+ real actions
Not just replies: it schedules, searches, executes, and automates with 50+ built-in tools and extensions.
Real memory
Persistent semantic memory across sessions, channels, and devices. Your context always present.
Multi-agent
Multiple coordinated agents working together on complex tasks, each with its own role.
Self-hosted
100% on your infrastructure. Your data never leaves you. Open-source (MIT): audit every line.

Extend it with skills & plugins from the Apilium Hub.

hub.apilium.com

Visit the Hub
How it works

One brain, your memory, your devices.

01

The brain

Mayros runs on a machine as the brain. It's where your agent's memory and logic live.

02

Your devices

Mobile, laptop, smart home… all connect to the brain. Messages come in from any app.

03

Act and log

Responds with context, executes actions with your tools, and leaves a signed log of each one.

Security

Agent power, with guardrails where it counts.

Isolated sandbox

Extensions cannot access your files, network, or system. They run isolated by default.

Code scanner

Every extension passes 5 automatic security checks before running.

Every action, logged

Prompt injection protection and a complete log of every decision — auditable.

Open-source · MIT

Free, auditable, and yours.

Mayros is open-source under the MIT license. Read it, contribute, and run it on your own infrastructure — no black boxes.

View on GitHub
Download

Install it on your machine in minutes.

Dependency-free desktop installers. Free and open-source.

Loading downloads...
FAQ

Common questions.

Is it free?

Yes. Mayros is open-source (MIT license) and free. You only pay for the API keys of the AI you choose — or use free local models with Ollama.

Which AI models are supported?

20+ providers: Claude, GPT, Gemini, Ollama (local models), Amazon Bedrock, and more. You choose and bring your own key.

Where does my data run?

100% self-hosted on your infrastructure. Mayros is the brain on your machine; nothing leaves you unless you decide it does.

Which channels does it support?

WhatsApp, Telegram, Slack, Discord, Teams, iMessage, Signal, and 20+ more. Your agent reaches you where you already are.

Your agent. Your devices. Your data.

Free, open-source, and self-hosted. Install in 5 minutes. No card, no lock-in.