Private AI on your Mac. June offers a unified, open-source workspace for chat, voice dictation, meeting notes, and a local agent, prioritizing user privacy by running locally by default. It routes model inference through a privacy-preserving, TEE-attested service, allowing users to verify its claims instead of trusting them blindly.

How It Works

June operates on a local-first principle, keeping sensitive data like notes, recordings, transcripts, files, and agent memory on the user's machine. When model inference is required, requests are routed through June API, an open-source, Trusted Execution Environment (TEE)-attested service. This service uses Intel TDX confidential VMs on Phala Cloud, ensuring provider keys remain server-side and routing to private models with zero data retention by default. The entire product is MIT licensed, and the production code is cryptographically verifiable.

Quick Start & Requirements

• Installation: Recommended via Homebrew (brew install --cask open-software-network/tap/june) or direct download for macOS. Windows builds are available but with limitations.
• Prerequisites: macOS 14 or later (Apple Silicon and Intel). For building from source: Node.js with pnpm 9 and a Rust toolchain.
• Resources: Free to start.
• Links: Website, Changelog, Verify, Telegram, X.

Highlighted Details

• Features: Includes AI chat, voice dictation with selectable styles, meeting notes (detects meetings, generates notes without a bot joining), a local agent built on the Hermes framework (sandboxed, requires approval for risky actions), and image generation.
• Model Choice: Users can select models from a live catalog, each labeled with its privacy tier, or use their own Venice API key.
• Verifiability: The production code running in the confidential VM is cryptographically verifiable against the source repository via OCI image labels and signed deployment tags.
• Extensibility: The local agent can be extended with skills, toolsets, and MCP servers.

Maintenance & Community

Development occurs in the open with near-daily releases. The primary community channel is Telegram (t.me/osjune), with updates also posted on X (@OpenSoftwareCo). Contributing guidelines are available in CONTRIBUTING.md and AGENTS.md.

Licensing & Compatibility

June is MIT licensed, permitting commercial use and integration into closed-source projects without copyleft obligations.

Limitations & Caveats

The primary development and target platform is macOS; Windows builds have reduced functionality, notably lacking global dictation paste and system audio capture. While June ensures zero data retention for its routed model calls, the privacy policies of the chosen upstream model providers are separate and may involve data retention.