MicroPaw

Personal assistant firmware for ESP32-S3. Runs GPT-5.6 Luna on the device with Telegram, memory, scheduled reminders, web search, Gmail and Google Calendar.

Supported hardware: ESP32-S3 N16R8 with 8 MB octal PSRAM and 16 MB flash.

Features

• LLM agent with tool calling (OpenAI, OpenRouter or custom Responses endpoint)
• Telegram and optional Instagram messaging
• Brave search, Wikipedia, arXiv, RSS/Atom readers
• Gmail (read, send, schedule, labels) and Google Calendar (list, create, update, delete)
• Voice note transcription and photo input
• Persistent memory (8 entries) and scheduled reminders (8 jobs)
• Morning briefing
• Optional SSD1306 OLED display
• OTA updates with signed releases

Install

Connect the ESP32-S3 over USB. On macOS or Linux:

curl -fsSL https://github.com/mahiatlinux/MicroPaw/releases/latest/download/install.sh | sh

In PowerShell:

irm https://github.com/mahiatlinux/MicroPaw/releases/latest/download/install.ps1 | iex

The installer needs Python 3.10 or newer. Set MICROPAW_PORT when more than one serial device is connected. If automatic reset fails, hold BOOT while the installer starts.

Configure

Create secrets/credentials.toml:

wifi_ssid = "NETWORK"
wifi_password = "PASSWORD"
telegram_token = "BOT_TOKEN"
owner_chat_id = "CHAT_ID"
llm_provider = "openrouter"
llm_api_key = "API_KEY"
llm_model = "openai/gpt-5.6-luna"

Push the TOML file over USB after the first flash. See docs/CONFIG.md for the full settings reference including Instagram, OLED, morning briefing, voice transcription and Google OAuth setup.

Build

idf.py build
idf.py -p PORT flash

Commands

Run /help on the device for available commands. See docs/COMMANDS.md for the full listing.