← Back to research
·5 min read·software

Lazzy

Lazzy is a floating AI agent for macOS that lives at your cursor. Detached interface, 800+ integrations, MCP support, 50+ AI models including local LLMs, and privacy-first architecture with local storage.

Key takeaways

  • Floating, detached AI interface for macOS — always available, never in the way
  • 800+ built-in integrations plus MCP support for external tools
  • Privacy-first — all conversations stored locally, zero data retention with hosted providers
  • Quick actions and workflow automation across your entire Mac

FAQ

How is Lazzy different from ChatGPT or Claude apps?

Lazzy is a floating agent that can orchestrate and automate apps across your Mac, not just a chat window. It works in the background and has 800+ integrations for taking action.

Is Lazzy private?

Yes — all conversations are stored locally on your device. Hosted AI providers are configured for zero data retention (no logs, no training).

What AI models does Lazzy support?

50+ models including Claude, GPT (4o through 5.2), Gemini, Grok, and DeepSeek, plus local models via Ollama and LM Studio — switchable on the fly for different tasks.

Overview

Lazzy is a floating AI agent for macOS designed to be "always available, never in the way."[1] Unlike traditional chat interfaces, Lazzy lives at your cursor as a detached, floating window — think Cursor's chat, but for your entire Mac.

The focus is on quick sessions, quick exits, and native macOS automation. Lazzy can orchestrate and automate apps across your Mac, working in the background without interruption or context switching.

As of June 2026, the product is live and labeled beta on its site.[1] Pricing has been published since the original profile, and the model lineup has expanded to 50+ models including local LLMs. Funding and user numbers are not publicly disclosed. Third-party directory DeepYard lists Lazzy at #70 among AI agents for 2026.[2]

Key Stats

MetricValue
PlatformmacOS (native), beta
InterfaceFloating, detached
Integrations800+ built-in
StorageLocal only
Models50+ (Claude, GPT, Gemini, Grok, DeepSeek, local via Ollama/LM Studio)
PricingFree ($0, 50 credits/mo) / Pro $20/mo or $5/mo annual + $0.01/credit
FundingNot publicly disclosed
Coming SooniOS Remote

Design Philosophy

Ergonomic by design — Not a bloated chat interface. Designed for quick interactions and quick exits. The floating window stays available but never blocks your workflow.

Detached operation — Works in the background. You're not switching to an app — the agent is embedded in your macOS experience.

Privacy-first — All data stored locally. Hosted AI providers configured for zero retention (no logs, no training, no leaks).

Features

  • Floating Interface — At your cursor, expandable when needed
  • Top AI Models — 50+ models (Claude, GPT 4o–5.2, Gemini, Grok, DeepSeek), switchable on the fly[1]
  • Local LLMs — Run models via Ollama or LM Studio, available even on the free tier[1]
  • Infinite Memory — Remembers preferences and project details
  • MCP & Tools — Connect external tools and data sources
  • 800+ Integrations — Built-in connections to apps and services
  • Quick Actions — Reusable workflows and prompts
  • Native Power — Automate complex workflows across macOS

Privacy Architecture

LayerProtection
StorageAll conversations local on device
ProvidersZero data retention configured
EncryptionEnd-to-end local privacy
TrainingYour data never used for training

Use Cases

Lazzy positions itself as a personal AI companion for:

  • Coding — Inline assistance while you work
  • Research — Quick lookups without context switching
  • Writing — Help at the cursor, not in another app
  • Planning — Project organization and task management
  • Automation — Cross-app workflows on macOS

Pricing

Pricing is now public (as of June 2026):[3]

PlanPriceIncludes
Free$0/mo50 monthly credits, limited models, quick actions, memories, integrations, local LLM support
Pro$20/mo (or $5/mo billed annually) + $0.01/credit overage500 monthly credits, all models, BYOK, screen sharing, early access, dedicated support

The Pro plan supports BYOK (bring your own API keys), which pairs with the privacy-first positioning — you can route inference through your own provider accounts or run fully local via Ollama/LM Studio.

Strengths

  • Native macOS — Purpose-built for Mac, not a cross-platform afterthought
  • Detached UX — Floating interface is genuinely different from chat apps
  • Privacy-First — Local storage, zero retention with providers
  • Quick Access — At your cursor, minimal friction
  • 800+ Integrations — Broad connectivity out of the box
  • MCP Support — Extensible with Model Context Protocol

Weaknesses / Considerations

  • macOS Only — No Windows, Linux, or mobile (iOS remote still "coming soon" as of June 2026)
  • New Product — Still in beta; less proven than established alternatives
  • No Public Traction — User and download numbers are not disclosed; little independent community discussion exists
  • Native Lock-in — Can't run on other platforms

Ideal User

  • macOS power users who want AI embedded in their workflow
  • Privacy-conscious users who want local-first storage
  • Quick interaction seekers who dislike full chat app context switches
  • Automation enthusiasts wanting cross-app macOS workflows

What Developers Say

No substantive independent user reviews or community discussion (Reddit, Hacker News, Product Hunt) of Lazzy were found as of June 2026 — the product is still in beta and visible adoption is thin. The closest third-party signal is a directory listing ranking it #70 among 2026 AI agents.[2] Note: "Lazy AI" (getlazy.ai) is an unrelated app-builder product and its reviews do not apply.

Bottom Line

Lazzy represents a different interaction model for personal AI — floating and detached rather than app-based. The "at your cursor" approach minimizes context switching and keeps the AI available without being intrusive.

The privacy-first architecture (local storage, zero provider retention, BYOK, local LLMs via Ollama/LM Studio) addresses a common concern with cloud AI. The 800+ built-in integrations and MCP support provide extensibility, and pricing is genuinely cheap — free tier or $5/mo annual Pro.

The main limitations are platform and proof: macOS only (iOS remote still pending), still in beta, and as of June 2026 there is essentially no independent community signal on whether it delivers. For Mac users who want an AI companion that's always present but never in the way, Lazzy offers a unique interaction model worth trying at the free tier. For cross-platform needs or proven tooling, look elsewhere.