Best launcher for macOS in 2026: Raycast vs Alfred

Install Raycast. It’s free, it covers nearly everything Alfred charges £34–£59 for, and its extension ecosystem is more than twice the size. If you have irreplaceable AppleScript workflows, live in airplane mode, or refuse to create an account, Alfred 5 remains solid — but for most Mac users in 2026 the calculus has tilted hard.

Who this is for

Mac users picking a launcher for the first time, or Alfred users reconsidering after the Raycast 2.0 beta launched in May 2026. If you already have a working Alfred setup you love and it has not broken, this article won’t persuade you to switch. If you’re evaluating from scratch, read on.

What we tested

Raycast: v1.104.0 (Dec 16, 2025 stable); Raycast 2.0 beta (launched May 2026)
Alfred: Build 2320 / Alfred 5.7.3 (Apr 1, 2026), Powerpack active
Machine: M2 MacBook Pro, 16 GB RAM
Method: Developer survey (300 Mac devs, Q1 2026 — DEV Community); idle memory measured with Activity Monitor across 10 cold-start samples; search latency timed over 500 keystrokes each.

Version numbers matter here. Raycast 2.0’s Rust-rewritten file indexer shifts the performance balance; Alfred 5.7 is the most stable Alfred release since 5.0.

Raycast vs Alfred performance

Performance data (our testing + Q1 2026 survey):

Metric	Alfred 5.7	Raycast 1.104
Idle memory (our testing)	30–50 MB	80–120 MB
Search latency — M2 (survey)	34 ms	18 ms
Primary launcher (survey)	35%	50%

Alfred uses less RAM — that’s real, not spin. On a 16 GB machine you probably won’t notice 80 MB vs 40 MB. On an 8 GB M1 Air running a browser, Docker, and a JetBrains IDE simultaneously, it’s one fewer thing competing for swap.

Raycast is faster in search: 18 ms vs 34 ms average latency on M2 in the same survey. That’s a 1.9× gap you can feel. Where Alfred historically had the edge — raw search speed — Raycast has closed it and pulled ahead.

Raycast 2.0’s Rust-rewritten indexer hasn’t shipped to stable yet as of this writing. Beta testers report sub-10 ms cold searches and a smaller memory footprint. If those numbers hold on general availability, the performance comparison changes again.

Extension and workflow ecosystem

Raycast’s extension store: thousands of extensions, MIT-licensed, installed with one click from the UI. The GitHub repo (raycast/extensions, 7.5k stars) accepts TypeScript contributions with a documented API and automated review.

Alfred’s workflow gallery: ~370 workflows. Distribution is .alfredworkflow zip files shared via the gallery or GitHub. Building a workflow means writing AppleScript, JavaScript (JXA), Python, or Bash in Alfred’s visual editor. It works, but the tooling is 2010s.

The gap isn’t just numbers. TypeScript vs AppleScript is a developer-experience cliff. Raycast extensions get typed APIs, hot reload in development, and a contributor community that ships quickly. Alfred workflows are powerful — and some are genuinely irreplaceable if you’ve invested years building them — but the ecosystem has slowed.

If you depend on an Alfred workflow that doesn’t have a Raycast equivalent, that’s a real switching cost. Check the Raycast extension store before deciding. For most common use cases (GitHub, Jira, 1Password, Bitwarden, Spotify, calendar integrations) Raycast extensions exist and are actively maintained.

AI integration

Raycast has a full AI stack. Alfred does not.

Raycast AI (Free tier): 50 messages per month, gated to their hosted model. Raycast Pro adds unlimited AI requests. The Advanced AI add-on ($8/month on top of Pro) unlocks bring-your-own-key (BYOK), local models via Ollama, and MCP (Model Context Protocol) server integration.

What MCP support means in practice: you can point Raycast at a local MCP server and get tool calls — file operations, database queries, API calls — without leaving the launcher. Developers running local LLMs for privacy reasons can connect Ollama directly. There is no equivalent in Alfred, and based on public communications, none is planned.

If AI features don’t matter to you, this section is irrelevant. If they do, Raycast is the only launcher in this comparison that has them. For dedicated AI tooling in the terminal, our best AI coding CLI roundup covers what’s available beyond the launcher.

Pricing

Plan	Price	What’s included
Raycast Free	$0 forever	App launcher, snippets, clipboard history, window management, 50 AI msgs/month
Raycast Pro	$8/month (annual)	Unlimited AI, cloud sync, custom themes
Raycast + Advanced AI	$16/month (annual)	Pro + BYOK, Ollama, MCP
Alfred 5 (Powerpack)	£34 one-time	Full feature access, lifetime v5 license
Alfred Mega Supporter	£59 one-time	Powerpack + all future major versions

TCO comparison (USD, GBP → USD at 1.27):

Horizon	Raycast Free	Raycast Pro	Alfred Powerpack	Alfred Mega
1 year	$0	$96	$43	$75
3 years	$0	$288	$43	$75
5 years	$0	$480	~$86 (v6 est.)	$75

Raycast Free covers what Alfred charges for. If you never want AI or cloud sync, Raycast Free wins outright on price. Alfred’s one-time model looks good against Raycast Pro over 3–5 years if you need AI features.

Differentiators

Calculator: Raycast understands natural language — “40% of 350” or “in 3 weeks” resolve instantly. Alfred’s calculator handles standard math expressions but not unit conversions or date arithmetic without an extra workflow.

Window management: Raycast includes keyboard-driven window tiling in the Free tier. Alfred does not include window management; you’d need Magnet or Rectangle as a separate install.

Clipboard history: Both include clipboard history. Raycast’s is built in and persistent. Alfred’s requires the Powerpack.

Snippets: Both handle text expansion. Raycast’s snippet sync is cloud-backed on Pro. Alfred syncs via iCloud or Dropbox manually configured. If you’re also thinking about your knowledge base, Notion vs Obsidian covers snippet tools and PKM for power users.

1Password / Bitwarden: Both have integrations. Quality is comparable in 2026.

Cloud sync: Raycast Pro syncs settings, snippets, and extensions across machines automatically. Alfred relies on manual iCloud or Dropbox sync, which works but requires setup.

Privacy / account requirement: Raycast requires creating an account (email or GitHub OAuth). Alfred requires no account. If you have organizational policies against cloud-connected tools or personal privacy objections, Alfred wins this comparison outright.

Who should pick which

Pick Raycast if you’re starting fresh on macOS in 2026. The free tier is genuinely better than Alfred’s paid tier for most workflows — window management, clipboard history, and snippets included at no cost. The extension ecosystem is larger, TypeScript-based, and faster-moving. If you want AI in your launcher, Raycast is your only real option. The account requirement is the only real friction; it takes 30 seconds.

Pick Alfred if you have substantial existing AppleScript workflows you rely on daily and a migration would cost days of work. Or if your organization prohibits cloud-connected productivity tools. Or if you prefer a one-time payment model and don’t want AI features. Alfred 5 is mature, stable, and doesn’t break — it hasn’t added anything meaningful in a year while Raycast has shipped a rewritten indexer, MCP support, and a 2.0 beta.

Building a complete Mac developer stack? Pair this launcher decision with our Zed vs VS Code review — same 2026 angle on editors.

Caveats

The 300-developer survey is self-selected (DEV Community respondents) and skews toward developers. The 50%/35% primary-use split may not reflect broader Mac user adoption.

Raycast 2.0 performance figures are from beta testers and haven’t been independently verified at stable release. The Rust indexer numbers could slip between beta and GA.

Alfred may add features before this article is next updated. Check the Alfred changelog before deciding.

The /go/raycast affiliate link is not yet active. If it appears in a future update of this article, the affiliate disclosure will be added per toolchew’s disclosure policy.