Skip to content

Repository files navigation

BoBe

A local-first proactive AI companion for macOS
BoBeBot.com

License macOS 15+ Apple Silicon


BoBe lives on your desktop as a transparent overlay. It watches what you're working on, builds memories over time, tracks your goals, and reaches out when it thinks it can help — like a thoughtful colleague who actually pays attention.

BoBe owns its data and runtime locally by default. You choose whether inference uses your GitHub Copilot subscription or a local Ollama model; only the context needed for a cloud answer leaves the Mac.

What BoBe Does

Observes your work Periodic screen captures analyzed by a vision model to understand context
Remembers you Maintains a local memory document; the agent can append durable facts through a bounded daemon tool
Tracks your goals Keeps human-readable goals, creates them after confirmation, and uses active goals proactively
Reaches out proactively A decision engine evaluates when help is valuable — not a chatbot waiting for input
Respects your flow Cooldown logic and engagement awareness prevent interruptions
Uses tools Read-only file access plus extensible MCP server integration
Customizable personality Soul documents shape how BoBe communicates

Supported Platforms

Platform Architecture Minimum Version Status
macOS Apple Silicon (arm64) macOS 15 Sequoia ✅ Supported
Ubuntu daemon x86_64 🧭 Compatibility target; no packaged release or continuous validation

Mobile and physical-device surfaces are future platform directions, not shipped applications. See docs/platform-architecture.md.

Quick Start

Option A: Download a Release

Download the latest BoBe.dmg from the Releases page, drag BoBe to Applications, and launch it.

Option B: Build from Source

git clone https://github.com/athatheo/BoBe.git
cd BoBe
just run

Requires macOS 15+, Rust 1.97.1+, Xcode 16+, and just. See CONTRIBUTING.md for full prerequisites.

First Launch

On first launch, BoBe's eight-stage setup wizard walks you through:

  1. Choose your AI — GitHub Copilot cloud or local Ollama
  2. Authenticate or install — sign in through BoBe's signed Copilot CLI helper, or let BoBe install Ollama and models
  3. Personalize — tell BoBe your name and optionally create a first goal
  4. Choose proactivity — tune how often BoBe observes and reaches out
  5. Grant permissions — screen awareness is optional
  6. Prepare voice — install the speech models used by the Mac

After setup, BoBe appears as a floating overlay on your desktop with a menu bar icon.

LLM Providers

Engine Description
GitHub Copilot Cloud inference using the user's Copilot subscription and CLI authentication.
Ollama Local inference. BoBe can install the runtime and selected models.

Configuration

All settings are configurable through BoBe's settings panel (click the menu bar icon → Settings). Settings persist to ~/.bobe/config.toml.

Environment variable overrides are available for advanced use:

BOBE_ENGINE__ENGINE=local
BOBE_CAPTURE__ENABLED=false
BOBE_CAPTURE__INTERVAL_SECONDS=30

Data is stored under ~/.bobe/ by default. Set BOBE_DATA_DIR to relocate the canonical data root; BOBE_DATABASE__URL remains available as an explicit SQLite override.

Security

BoBe handles sensitive data including screen captures, personal context, and integration secrets. See SECURITY.md for our vulnerability reporting policy.

Key security properties:

  • Loopback by default — the daemon binds to 127.0.0.1 unless remote mode is deliberately configured
  • Secrets stored in macOS Keychain and handled in-memory with the secrecy crate
  • Arbitrary file writes and shell commands are denied; memory/goal mutations use bounded daemon-owned tools
  • MCP commands validated against a configurable blocklist
  • CORS locked to localhost origins
  • Remote/reverse-proxy access must configure an API bearer token and TLS certificate/key; adding a public allowed host without both is rejected at startup, including when the daemon itself binds to loopback behind a proxy

Contributing

We welcome contributions! See CONTRIBUTING.md for development setup, architecture, coding conventions, and the full project structure.

Code of Conduct

This project follows the Contributor Covenant Code of Conduct. By participating, you are expected to uphold this code.

License

MIT License

About

A local-first proactive AI companion for macOS

Resources

Code of conduct

Contributing

Security policy

Stars

8 stars

Watchers

1 watching

Forks

Releases

Packages

Contributors

Languages