Private AI on your Mac. Chat, dictation, meeting notes, and a local agent in one open source workspace.
# Add to your Claude Code skills
git clone https://github.com/open-software-network/os-clovyGuides for using ai agents skills like os-clovy.
Last scanned: 8/13/2026
{
"issues": [],
"status": "PASSED",
"scannedAt": "2026-08-13T05:40:08.549Z",
"npmAuditRan": false,
"pipAuditRan": true,
"promptInjectionRan": true
}Most AI apps ask you to hand over your most sensitive data and trust them with it. Every prompt, file, and meeting reveals something about you, and a cloud agent with that reach is a remote company's window into your work.
Clovy is built the other way around. The app and the agent run on your Mac. Notes, recordings, transcripts, files, sessions, and agent memory stay on your machine by default. When Clovy needs model inference, the request goes through Clovy API, an open source, TEE-attested service that keeps provider keys server-side and routes to private models with zero data retention by default. You do not have to take any of this on faith: the entire product is MIT licensed, and the exact code serving production is cryptographically verifiable.
Local by default. App state, recordings, transcripts, and agent memory live on your machine. The agent runs locally inside a macOS write-jail unless you opt a session out.
Private models. Model calls default to private Venice models with zero data retention: nothing stored, no training. Anonymized third-party models are opt-in, and those providers may retain what they receive under their own policies.
Minimal retention. Open Software's services store account, login, and billing records. Prompts, audio, transcripts, and files are not among them.
Verifiable, not promised. Clovy and Clovy API can be inspected independently of any model routing service or Chat. The desktop releases are signed and notarized, and their source is public. Clovy API runs in an Intel TDX confidential VM on Phala Cloud and publishes three useful anchors:
org.opencontainers.image.revision label.build-clovy-api.yml
publishes ghcr.io/open-software-network/clovy-api;
deploys pin immutable per-commit tags recorded as signed deploy/<env>/<sha> git tags.Every deployment serves a self-contained walkthrough at
/verify. This evidence describes
Clovy API only. The Open Software API and Chat publish their own source and
runtime evidence, and Clovy does not need to pin their releases. Model privacy
remains explicit provider evidence, which is why zero-retention private
routing is the default.
Clovy runs on macOS 14 or later, Apple Silicon and Intel. Releases are signed,
notarized, and auto-updating, with stable and rc channels switchable
in-app. It is free to start.
If you use Homebrew, this is the recommended way to install:
brew install --cask open-software-network/tap/june
Windows builds cover the app shell, sign-in, microphone recording, notes, and the bundled agent runtime, but not global dictation paste, system audio capture, or the macOS sandbox. macOS is the primary target.
This repo contains the full product: the desktop app and the service that powers its metered AI calls.
src/ React and TypeScript frontend
src-tauri/ Tauri v2 Rust desktop backend and native helpers
clovy-api/ Clovy API: models, transcription, generation, and billing
docs/ Architecture notes, ADRs, subsystem guides, and runbooks
spec/ Enforceable coding rules
specs/ Feature specs, plans, and validation notes
The desktop app never holds provider or OS Accounts App API keys; those live only in Clovy API. Start with docs/index.md for the full doc map, CONTEXT.md for the domain glossary, and AGENTS.md for the contributor guide.
You need Node.js with pnpm 11 and a Rust toolchain. The exact pnpm version is
pinned in package.json.
git clone https://github.com/open-software-network/os-clovy
cd os-clovy
cp .env.example .env
cp clovy-api/.env.example clovy-api/.env
# Edit clovy-api/.env and set CLOVY__UPSTREAMS__VENICE__API_KEY.
pnpm install
pnpm tauri:dev
The example env files default to open source local mode: no OS Accounts login,
no billing, and a local Clovy API authenticated by a shared bearer token.
Provider keys belong only in clovy-api/.env, never in the root desktop
.env. A Venice API key is enough for transcription, generation, and
dictation cleanup.
See docs/development.md for the day-to-day development guide (ports, onboarding replay, local data, permissions, agent skills, and test commands) and docs/configuration.md for the full configuration reference, including exposing your own models and running against OS Accounts.
Clovy ships near-daily releases and development happens in the open.
Clovy is also the canonical technical name for repository-controlled packages, crates, environment variables, workflows, and release artifacts. Released June-era inputs and rollback outputs remain compatibility aliases, while the installed bundle, executable, updater, and permission identities stay stable. See ADR-0055.
pnpm check # lint and format (Biome)
pnpm typecheck
pnpm test # frontend (Vitest)
pnpm test:rust # desktop Rust
pnpm test:clovy-api # Clovy API
make verify mirrors CI. Start with CONTRIBUTING.md, then
AGENTS.md for the full contributor guide; the enforceable UI
rules live in spec/. Report bugs through GitHub issues, and
report security vulnerabilities privately per SECURITY.md.
Clovy is MIT licensed. See LICENSE. Bundled third-part
os-clovy is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by open-software-network. Private AI on your Mac. Chat, dictation, meeting notes, and a local agent in one open source workspace. It has 345 GitHub stars.
Yes. os-clovy passed SkillsLLM's automated security scan — a dependency vulnerability audit plus prompt-injection heuristics — with no high-severity issues. You can read the full report in the Security Report section on this page.
Clone the repository with "git clone https://github.com/open-software-network/os-clovy" and add it to your Claude Code skills directory (see the Installation section above).
os-clovy is primarily written in Rust. It is open-source under open-software-network on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh os-clovy against similar tools.
No comments yet. Be the first to share your thoughts!