mikeyobrien/rho

What it solves

rho is an always-on personal AI operator designed to move beyond stateless chat interfaces. It solves the problem of AI tools forgetting context between sessions by providing persistent memory and the ability to proactively check in with the user on a schedule rather than only responding to prompts.

How it works

Built on the pi coding agent, rho runs as a background daemon on macOS, Linux, and Android. It utilizes a local-first state model where memory and configuration are stored on the user's machine. The system consists of several core components:

• Heartbeat: A scheduler that triggers autonomous check-ins (default every 30 minutes).
• Brain: An append-only structured memory file (brain.jsonl) for durable context.
• Vault: A markdown-based knowledge graph for captured information.
• Multi-surface Control: Users can interact with the agent via a CLI, a built-in lightweight Web UI, Telegram, or a dedicated agent email address.

Who it’s for

It is for users who want a persistent AI assistant that remembers their preferences and tasks across sessions, developers who need a memory-backed coding copilot, and those who prefer local-first AI control with their own model providers.

Highlights

• Persistent Memory: Durable context that can be inspected, searched, and edited via the /brain command.
• Proactive Operation: Scheduled heartbeats for reminders and task management.
• Local-First Security: Memory and configuration stay on the user's local machine.
• Multi-Channel Access: Integrated support for Telegram and email (name@rhobot.dev).
• Cross-Platform: Native support for macOS, Linux, and Android (via Termux or a native wrapper), with SSH access for iOS.