ownpilot/OwnPilot

What it solves

OwnPilot is a privacy-first, self-hosted personal AI assistant platform. It solves the problem of relying on centralized AI services by allowing users to maintain full control over their data while accessing a wide array of AI capabilities, including autonomous agents, tool orchestration, and integration with personal productivity data.

How it works

The platform operates as a unified gateway that connects a React-based web UI and messaging channels (Telegram, WhatsApp) to a core AI engine. This engine supports multiple AI providers (both cloud-based and local via Ollama or vLLM) and orchestrates a vast library of over 250 built-in tools. It employs a "Meta-tool Proxy" to keep LLM context windows clean by only sending a few core tools to the model, which then dynamically discovers other tools as needed.

Who it’s for

It is designed for users who want a highly customizable, autonomous AI assistant that can manage personal data, automate web browsing, control IoT devices, and execute code in a sandboxed environment, all while keeping their data on their own hardware.

Highlights

• Claw Agents: Autonomous runtimes that combine reasoning, isolated workspaces, and persistent directive files to complete complex tasks.
• Soul Agents: Agents with distinct personalities, missions, and "heartbeat" lifecycles for periodic autonomous operation.
• Extensive Tooling: Over 250 built-in tools and support for the Model Context Protocol (MCP) to connect to external servers.
• Multi-Channel Access: Integration with Telegram and WhatsApp for interacting with the assistant from mobile devices.
• IoT & Edge Control: Built-in MQTT integration to monitor and control Raspberry Pi, ESP32, and other edge devices.
• Privacy & Security: Features AES-256-GCM encryption, PII redaction, and Docker-based sandboxing for code execution.