Best AI App for Mac: Raycast 2.8 Still Drives the Screen

Best AI App for Mac: Raycast 2.8 Still Drives the Screen

Last updated on

Almost every AI app on the Mac now calls itself an agent. Most of them are not. The word has a specific meaning: software that takes an instruction, decides on a sequence of steps, and executes those steps against real systems. The gap between apps that do that on macOS and apps that only appear to is wide enough to matter before you install anything.

This page is only about the ones that act. The ranked list of Mac AI assistants of every kind lives in 10 best AI assistants for Mac. If the job was a keyboard launcher rather than an agent, that is Raycast vs Alfred on Mac. Raycast’s current Mac download is v2.8.2.0 (arm64 DMG Raycast_2.8.2.0_d6aec411b0_arm64.dmg, raycast.com/download, 2026-10-16, Content-Length 145,421,420). raycast.com/new still requires macOS Tahoe and Apple Silicon. v2.8 (page date October 16, 2026) is a speed release: File Search up to 6× faster, extensions about 4× faster (up to 30×), about 20% less memory at launch. Send Element to AI and Send Window to AI still pass what is on screen into chat. That is screen context. It is not Finder control.

The 2026 line moved twice. Siri AI shipped with macOS 27 on 2026-09-14 (not on Tahoe). Claude computer use is in beta on Pro and Max, not a research preview (help center dateModified 2026-10-14). Neither one calls Finder, Mail, or EventKit as a first-class tool. Product definition, including the three model paths: What Is Dottie Desktop (1,623 downloads as of 2026-10-16, desktop 2026.9.14). Why the agent itself has no client fee: Why I Made My Mac AI Assistant Free Forever.

TL;DR

Agent vs Assistant: The Test That Matters

Marketing copy will not settle this. One question will:

After the AI finishes, is anything different on my Mac that I did not do myself?

If the answer is no — you got text back, and you were the one who acted on it — that is an assistant, however good the text was. If the answer is yes — a file moved, an app opened, a message sent, a setting changed — that is an agent.

Three things have to be true for the second answer to be possible:

  1. Perception — it can tell what is currently on your Mac.
  2. Action — it has a way to change that state.
  3. Iteration — it can check the result and adjust, rather than firing one command and stopping.

Missing any one of the three and you have a chatbot with extra steps. Most “AI agent for Mac” listings fail on #2.

The Four Tiers of Mac AI Agent

Every Mac option in 2026 sits in one of these. The tier tells you more about what an app can do for you than its feature list does.

TierHow it actsWhat it can reachExamples
Native API agentCalls macOS APIs and shell directlyYour whole MacDottie
Screen-driving agentScreenshots, then simulates clicks and keystrokesWhatever is visible on screenClaude Cowork
Sandboxed agentCalls functions the host app exposesOnly that app’s integrationsRaycast AI Extensions
Remote / computer-useCloud plugins, plus click/type on the desktopGUI and vendor SaaS, not macOS APIsChatGPT Work

The tiers are not a quality ranking. A remote agent is genuinely better for “book me a flight” than anything running locally, because it can grind through a checkout flow for ten minutes without borrowing your cursor. They are a reach ranking. Pick the tier that matches where your task lives.

A native multi-model chat client is not in this table. BoltAI talks to many models in one Mac window (pay once, BYOK, local runners). It does not call Finder or EventKit. Keep it if the job is chat. Use an agent if the job is the Mac.

Native API Agents

Dottie — Full System Access, Free

Dottie is a free, open-source AI agent for macOS that calls system APIs rather than driving the interface. It ships with 134 built-in system tools — files, apps, system settings, messages, screen reading — and supports 3,800+ local models through MLX, so the whole loop can run offline on Apple Silicon.

Why the native approach matters in practice: an agent that clicks through a UI is blocked by anything the UI does not expose, breaks when a window moves, and cannot work while you are using the machine. An agent calling APIs has none of those failure modes. The trade-off is the mirror image — it can only reach what it has a tool for, whereas a screen-driver can attempt anything a human could.

  • Voice-first, with a “Hey Dottie” wake word, barge-in, and text-to-speech
  • Runs fully offline with local models; nothing leaves the Mac
  • Free and open source; local models, BYOK, or optional Dottie Pro at $20/month
  • Multi-provider — bring your own key for OpenAI, Anthropic, xAI, or Cerebras
  • Requires macOS 15.4+ on Apple Silicon
  • Cumulative downloads past 1,600 (1,623 as of 2026-10-16, api.dottie.ai). Latest desktop 2026.9.14. Local runner: Ollama 0.40.0 runs supported architectures on MLX by default on Apple Silicon. GitHub Latest is 0.40.5 (published 2026-10-14). The notes add Nemotron 3.5 Super, updated llama.cpp, and updated MLX. They do not add an Apple Silicon tok/s number. Setup: run local AI models on Mac. Why that client has no fee: Why I Made My Mac AI Assistant Free Forever.

Best for: recurring work on your own machine — file wrangling, app control, hands-free operation while your hands are busy.

Download Dottie

Screen-Driving Agents

Claude Cowork — Sees the Screen, Simulates the Human

Claude’s computer use takes screenshots, reasons about what is on them, and clicks and types. That gives it a reach no API-based tool can match: if a human could do it by looking at the screen, it can attempt it, including in apps that expose no automation surface at all.

The help center (re-checked 2026-10-16, dateModified 2026-10-14) is specific. Computer use is still in beta on Pro and Max, in Claude Desktop on macOS and Windows. It is not a research preview. Cowork and chat are one Claude on those plans. Claude prefers connectors, then the built-in browser, then screen interaction. On macOS 15 or later it works in background windows and does not take your pointer by default. It does not call Finder, Mail, or EventKit. Anthropic notes screen interaction is slower than a connector and complex tasks sometimes need a second try. Screenshots of whatever is visible go to Anthropic.

Best for: one-off tasks in apps that have no scripting interface, where reach matters more than speed.

See Claude Cowork vs Dottie for a task-by-task comparison of the two approaches, and Dottie vs Claude for the broader product difference.

Sandboxed Agents

Raycast AI — Real Actions, Bounded Scope

Raycast AI Extensions let the assistant call functions that installed extensions expose, so “create a Linear issue from this” genuinely creates the issue. That is agentic by the test above: state changed, and you did not change it.

The boundary is the extension catalogue. Raycast can act on the services someone has written an extension for. Current Mac download is v2.8.2.0 (Content-Length 145,421,420, 2026-10-16): macOS Tahoe and Apple Silicon. v2.8 (page date October 16, 2026) is a speed release: File Search up to 6× faster, extensions about 4× faster (up to 30×), about 20% less memory at launch. Send Element to AI and Send Window to AI still pass what is on screen into chat. The chat settings list names Terminal, Finder, and Automations among extensions whose tools the chat uses. Naming Finder in that list is not a Finder API. v2.7 (October 7, 2026) ran AI Chat, Notes, and Settings in their own processes and added those send-to-AI actions. v2.5 (2026-09-25) lets Store extensions supply models in AI Chat, including Ollama and LM Studio. Raycast Pro still starts at $8/month (page checked 2026-10-16). v2.2 gated Custom Providers, Local Models, and BYOK behind Pro. Do not read “extension models” as free local AI. Sequoia stays on v1. The two-bar duel: Raycast vs Alfred on Mac. Launcher field: Raycast alternatives for Mac (title still names v2.8.0; live download is v2.8.2.0).

Best for: keyboard-driven workflows across SaaS tools you already have extensions for.

Remote / Computer-Use Agents

ChatGPT Work — Computer Use, Not Native APIs

ChatGPT Work replaced Agent Mode in July 2026. On the Mac desktop app it can use local files, cloud plugins (Drive, Slack, mail, calendar), a built-in browser, and computer use (click, type, move files). That is real agency. It is also the same class of control as Claude Cowork: pixels and simulated input, not EventKit / Finder / Mail APIs. Work with Apps reads a supported allowlist. Nothing runs offline.

Best for: long OpenAI-plugin jobs and GUI tasks you would rather not babysit. Full split: Dottie vs ChatGPT Desktop.

What About Apple Intelligence?

Apple Intelligence is not in the four-tier table because Writing Tools, notification summaries, and on-device language features are still assistants. Siri AI is the part that now acts. It shipped with macOS 27 Golden Gate on 2026-09-14: Spotlight chat, personal context across messages, mail, and photos, app actions, a Siri app, and Visual Intelligence on Shift-Command-Space. Ars Technica the same day: an M-series Mac is required, and Siri can interact with an app when that app’s developer added support. It is not on Tahoe. The support note does not add a third-party model picker or an open tool list you can extend.

Treat Siri AI as a more capable system assistant. Treat Dottie as the native API agent on Sequoia 15.4+ and on Golden Gate. Voice-only split: Siri alternative for Mac. Broader field: Apple Intelligence alternatives for Mac.

Picking a Tier by Task

What you want doneTier you needPick
Rename, sort, or move local filesNative APIDottie
Operate a Mac app with no scripting supportScreen-drivingClaude computer use (beta, Pro/Max)
Run a task hands-free by voiceNative APIDottie
Fire an action into Linear, Notion, or SlackSandboxedRaycast AI
Spend twenty minutes filling in a web formComputer useChatGPT Work
Keep every byte on the machineNative API, local modelsDottie
Apple house: HomeKit, calls, Writing ToolsBuilt-in assistantSiri / Siri AI on macOS 27

Privacy: Where the Agent Runs Decides Everything

For agents the privacy question is sharper than for chat, because agents see more than what you typed — they see your screen, your files, and the results of their own actions.

  1. Fully local — Dottie with local models. The prompt, the screen contents, and the tool results never leave the Mac.
  2. Screen sent to a vendor — Claude Cowork. Every step ships a screenshot of whatever is visible, including anything you forgot to close.
  3. Cloud by default — ChatGPT Work, and Raycast AI unless you point AI Chat at a local model. Prompts and results otherwise go to someone else’s servers.

For regulated work — health, legal, client-confidential — tier 1 is the only one that survives a review. Running local AI models on your Mac covers the setup. Ollama 0.40.0 runs supported architectures on MLX by default on Apple Silicon. GitHub Latest is 0.40.5 (published 2026-10-14). No Apple Silicon tok/s in those notes. Ollama vs LM Studio compares the two runners.

Frequently Asked Questions

What is the best AI app for Mac?

It depends on the job. For changing the Mac through native APIs, including offline on Apple Silicon, Dottie is the free agent (macOS 15.4+). For keyboard actions inside extensions, Raycast 2.8 still needs Tahoe and Apple Silicon (Mac download v2.8.2.0 on 2026-10-16, Content-Length 145,421,420). Send Element to AI hands a hovered UI element to chat. That is screen context, not Finder. The v2.8 changelog, dated October 16, 2026, is a speed release. For clicking whatever is on screen, Claude computer use is still beta on Pro and Max (help center re-checked 2026-10-16). Siri AI ships with macOS 27, not Tahoe. One app does not cover all four.

What is an AI agent for Mac?

An AI agent for Mac is software that takes an instruction, plans a sequence of steps, and executes those steps against your machine — opening apps, moving files, changing settings — then checks the result and adjusts. The test that separates an agent from an assistant is whether anything on your Mac is different after it finishes that you did not change yourself. Most Mac AI apps marketed as agents only return text.

What is the difference between an AI agent and an AI assistant?

An assistant answers questions and produces text; you remain the one who acts on it. An agent perceives the current state of your system, has a mechanism to change that state, and iterates until the task is done. ChatGPT answering 'how do I batch-rename these files' is an assistant. Software that renames them is an agent.

Which AI agents can actually control a Mac?

Dottie controls macOS through 134 native system tools, calling APIs directly. Claude computer use (still beta on Pro and Max, re-checked 2026-10-16) and ChatGPT Work control it by taking screenshots or simulating clicks and keystrokes. On macOS 15 or later, Claude works in background windows and does not take the pointer by default. Raycast 2.8 can trigger extension actions and can send a hovered element or a window into AI Chat. The settings list names Terminal, Finder, and Automations as extensions. That is not a Finder API. Siri AI on macOS 27 can act across apps whose developers added support. ChatGPT Agent Mode is gone; OpenAI points users to ChatGPT Work.

Did Siri AI make Apple Intelligence an agent?

Siri AI shipped with macOS 27 Golden Gate on 2026-09-14. Apple's support note: Spotlight chat, personal context, app actions, a Siri app, and Visual Intelligence on Shift-Command-Space. Ars the same day: M-series required, and app actions need developer support. That is a more capable system assistant. It is not on Tahoe, and the note does not publish an open tool list or a third-party model picker. Native API control on Sequoia 15.4+ is still Dottie.

Is there a free AI agent for Mac?

Dottie is free and open source and runs local models at no cost on Apple Silicon. Optional Dottie Pro at $20 per month is managed model access, not a lock on the app. Claude Cowork requires a paid Claude plan, Raycast Pro starts at $8/month billed annually, and ChatGPT Work is limited on the free desktop tier.

Can an AI agent run offline on a Mac?

Dottie runs the full agent loop offline using local models through MLX on Apple Silicon, so prompts, screen contents, and tool results never leave the machine. Claude computer use and ChatGPT Work need a network. Raycast can point AI Chat at a local model (Ollama or LM Studio via an extension since v2.5). v2.8 (October 16, 2026) keeps local-model images on-device when Cloud Sync or AI Chat History sync is off. Cloud Raycast models and most SaaS extension actions still need a network. Ollama 0.40.0 runs supported architectures on MLX by default on Apple Silicon. GitHub Latest is 0.40.5 (published 2026-10-14) and does not add an Apple Silicon tok/s number.

Is a native agent better than a screen-driving agent?

Neither is strictly better. They trade reach against reliability. A native API agent is faster and does not break when a window moves, but it can only reach what it has a tool for. A screen-driving agent can attempt anything a human could do by looking at the screen, including in apps with no automation support. Claude computer use is slower than a connector, sometimes needs a second try, and on macOS 15 or later stays in background windows by default. Screenshots of whatever is visible go to the vendor.

Is Raycast AI an agent for Mac?

Inside its extension sandbox, yes: calling an extension function changes state you did not change yourself. It is not a system agent. v2.8 (October 16, 2026) can send a hovered element or a whole window into AI Chat, and the chat settings list names Terminal, Finder, and Automations as extensions whose tools the chat uses. That is still the extension catalogue, not Finder or EventKit. Current Mac download is v2.8.2.0 (checked 2026-10-16, Content-Length 145,421,420) and needs macOS Tahoe plus Apple Silicon. v2.5 lets extensions supply AI Chat models, including Ollama. If you wanted a launcher rather than an agent, read Raycast vs Alfred on Mac.

← Back to Blog

More from the blog