Meta Muse features: 7 things it does (cross-app memory + booking)
Meta Muse features (Sept 8 launch): cross-app memory across Facebook/Instagram/WhatsApp, proactive reminders, agentic booking, calendar sync, photo search. Privacy opt-out + vs ChatGPT comparison.
Meta launched Muse, the world's first personal AI agent built for everyone, on September 8, 2026 — six days after the developer-focused Spark 1.3 model. Muse runs inside Facebook, Instagram, WhatsApp, and Meta AI and goes far beyond a chatbot: it remembers context across apps, surfaces proactive reminders without being asked, and executes multi-step tasks like booking flights and hotels. The launch is the most consequential Meta product release of 2026 because it puts a persistent AI agent in front of 3.2 billion monthly active users across the Meta family of apps. Chief AI Officer Alexandr Wang announced the launch and positioned Muse as the model family that delivers "personal superintelligence."
What is the Meta Muse personal agent?
The Meta Muse personal agent is a cross-app AI assistant that lives inside Facebook, Instagram, WhatsApp, and Meta AI. Unlike ChatGPT or Claude (which wait for a prompt), Muse is proactive — it reads your messages, calendar, and behavior across Meta's apps and surfaces reminders, suggestions, and actions based on context you didn't explicitly share.
The agent launched globally on September 8, 2026. Rollout is staged by app:
| App | Muse integration | As of |
|---|---|---|
| Meta AI (standalone) | Full agent + chat | Sept 8, 2026 |
| Agent + voice notes | Sept 10, 2026 | |
| DMs + Reels search | Sept 12, 2026 | |
| Messenger + Events | Sept 15, 2026 |
By Sept 22, 2026, Muse is live in all four apps globally. The rollout is server-side — users do not need to download an update, but they must opt in once via Settings → Privacy → Muse.
What can the Meta Muse personal agent do?
Muse can read context across all four Meta apps, remember it indefinitely, and take actions on your behalf — surfacing reminders, completing purchases, booking travel, and answering questions grounded in your personal data. It is the first personal-MAI assistant shipped at Meta's scale.
| Capability | Description | Example |
|---|---|---|
| Cross-app memory | Remembers context from FB events, IG DMs, WhatsApp chats | "What did Sarah say about the trip?" |
| Proactive reminders | Surfaces reminders without being asked | "Your package ships tomorrow" |
| Multi-step agentic tasks | Books flights, hotels, restaurants across providers | "Plan a trip to Tokyo next month" |
| Voice + text | Works in WhatsApp voice notes, voice calls, text | Dictate a reminder while driving |
| Reels / Marketplace search | Natural-language search inside Instagram / FB | "Show me hiking boots under $100" |
| Calendar integration | Pulls events from Messenger chats into reminders | "Sarah's birthday is Friday" |
The booking capability is the most novel: Muse integrates with Booking.com, Airbnb, OpenTable, Expedia, and major airlines via Meta's existing commerce partnerships. When you say "Plan a trip to Tokyo next month," Muse searches flights and hotels, shows you 3 ranked options, and waits for confirmation before booking.
How does Muse differ from ChatGPT or Claude?
ChatGPT and Claude are chatbots that wait for prompts; Muse is a proactive agent that lives across your daily apps. The architectural difference is what data Muse can see and what actions it can take.
| Dimension | Meta Muse | OpenAI ChatGPT | Anthropic Claude |
|---|---|---|---|
| Where it lives | FB / IG / WhatsApp / Meta AI | Web + iOS / Android app | Web + iOS / Android app |
| Reaches you proactively? | Yes — surfaces reminders | No — waits for prompt | No — waits for prompt |
| Sees your messages? | Yes — across 4 Meta apps | Only what you paste in | Only what you paste in |
| Books travel autonomously? | Yes — with confirmation | Operator (separate product) | Computer use (beta) |
| Personal context | Cross-app memory by default | Custom GPTs (manual setup) | Projects (per-conversation) |
| Privacy opt-out | Settings → Privacy → Muse off | Account delete / training opt-out | Account delete / training opt-out |
Muse trades user-data privacy for capability. The cross-app memory is what makes the agent useful; turning it off disables most proactive features.
Is the Meta Muse agent free?
Yes — Muse is free for all users with a Meta account. There is no Muse Plus or Muse Pro tier. The cost to Meta is significant: each Muse interaction costs roughly 3-8 cents in compute (per Meta's published Spark 1.3 pricing), and Meta expects 4-6 daily Muse interactions per active user by end of 2026. Meta is monetizing Muse indirectly via ads — the agent's booking recommendations favor Meta-commerce partners, and the agent surfaces "Sponsored" reminders in the Facebook feed.
This is the same ad-funded free model that powers Facebook and Instagram. The agent is free because the data and the engagement are the business model.
How does Meta Muse handle privacy?
Muse reads your messages and behavior across Facebook, Instagram, WhatsApp, and Meta AI to build a per-user agent memory — but does not use that data to train the underlying Spark 1.3 model. Meta has published a privacy white paper (Sept 8, 2026) distinguishing:
- Agent memory (per-user, stored in your Meta account): used to power Muse's responses to you. Not shared with other users, not used for ad targeting, deletable from Settings → Privacy → Muse → Delete Memory.
- Model training (Spark 1.3 / Glimmer weights): trained on public web data + licensed datasets. Not trained on user messages or agent memory.
- Ad targeting (separate from Muse): Facebook / Instagram ads still use your engagement data, but Muse suggestions are not ad-injected unless marked "Sponsored."
Three user controls:
- Disable Muse entirely: Settings → Privacy → Muse → Off. Disables all agent features across all four apps.
- Delete agent memory: Settings → Privacy → Muse → Delete Memory. Removes the per-user memory store. Existing proactive features stop working until memory rebuilds.
- Per-app control: You can disable Muse in one app (e.g., Instagram) and keep it in another (e.g., WhatsApp). Granularity is per-app.
Important caveat: Meta's privacy policy still allows the company to access your messages for content moderation, safety review, and legal compliance. Muse's privacy is layered on top of Meta's existing data-access posture — it doesn't restrict Meta's internal access.
What is "personal superintelligence" and why is Alexandr Wang using that term?
"Personal superintelligence" is Alexandr Wang's framing for AI that knows you deeply enough to act in your interest without explicit prompts — surfacing reminders, completing tasks, and making decisions grounded in your preferences, history, and context. Wang, Meta's Chief AI Officer (formerly Scale AI CEO, joined mid-2025), has used the term in every Muse launch since August 2026.
The term is a deliberate parallel to OpenAI's "superintelligence" framing (Sam Altman, May 2024) and Anthropic's "powerful AI" framing (Dario Amodei, 2024). Wang's twist: Meta's superintelligence is personal — the model knows you, not the world. Critics argue this is positioning to justify the data-collection scope.
When did the Meta Muse agent launch?
| Date | Event |
|---|---|
| Aug 2026 | Muse Glimmer (open-weight model) released |
| Sept 2, 2026 | Muse Spark 1.3 (closed API model) released |
| Sept 8, 2026 | Muse personal agent announced at Meta Connect |
| Sept 10, 2026 | Muse available in WhatsApp |
| Sept 12, 2026 | Muse available in Instagram |
| Sept 15, 2026 | Muse available in Facebook |
| Q2-Q3 2027 (projected) | Watermelon (next-gen flagship model) |
Muse was announced at Meta Connect 2026, Wang's first keynote as Chief AI Officer.
Should I opt in to Muse or wait?
Opt in if: you already use FB / IG / WhatsApp daily, you're comfortable with Meta accessing your messages, and you want a proactive agent for travel / shopping / reminders.
Wait / opt out if: you have sensitive conversations on WhatsApp (medical, financial, legal), you don't want Meta to read your DMs even with the privacy controls, or you prefer reactive chatbots over proactive agents.
For most users, the right answer is opt in + delete memory weekly. Muse is the strongest consumer agent shipped as of September 2026, and the convenience is significant — but the data-access scope is the broadest of any consumer AI product.
Bottom line: Meta Muse is the first personal AI agent shipped at Meta's 3.2-billion-user scale. It goes beyond ChatGPT and Claude by being proactive, cross-app, and capable of executing real bookings. The trade-off is data access — Muse reads your messages across four apps by default. For users comfortable with that scope, Muse is the strongest consumer AI assistant available today.
For the underlying Spark 1.3 model (the engine powering Muse), see our Muse lineup FAQ. For a deeper comparison of Spark 1.3 vs GPT-5.6 vs Claude Opus 5, see our AI model benchmark deep-dive.
How does Muse compare to ChatGPT, Claude, and Gemini?
Muse is the most proactive of the four consumer agents, but the broadest in data access. ChatGPT is the most versatile reactive chatbot. Claude is best for long-context reasoning. Gemini is the only one with Workspace integration. Direct comparison:
| Capability | Meta Muse | ChatGPT Plus | Claude Pro | Gemini Advanced |
|---|---|---|---|---|
| Proactive reminders | ✅ Yes (core feature) | ❌ No (reactive only) | ❌ No | ⚠️ Limited (Workspace events) |
| Cross-app memory | ✅ FB + IG + WhatsApp + Meta AI | ⚠️ ChatGPT memory only | ⚠️ Project memory | ✅ Google Workspace + Android |
| Agentic booking (flights/hotels) | ✅ Booking.com, Airbnb, OpenTable, Expedia | ⚠️ Limited (Operator in US) | ⚠️ Computer use (beta) | ⚠️ Limited |
| Voice + text | ✅ WhatsApp voice notes | ✅ Voice mode | ✅ Voice (mobile) | ✅ Gemini Live |
| Image generation | ✅ Meta AI image gen | ✅ DALL-E 3 | ❌ No (text only) | ✅ Imagen 3 |
| Web browsing | ✅ (via Meta AI) | ✅ Search | ✅ Web search | ✅ Grounded in Google Search |
| Privacy: data scope | ⚠️ Reads FB/IG/WhatsApp DMs | ⚠️ Chat history + memory | ✅ No data training by default | ⚠️ Workspace data |
| Free tier | ✅ (with ads, no booking) | ✅ (GPT-4o mini, rate-limited) | ✅ (Sonnet, rate-limited) | ✅ (Gemini 1.5 Flash) |
| Paid price | Free (no Premium tier) | $20/mo | $20/mo | $20/mo |
| Best for | Cross-app personal automation | Versatile chat + tools | Long reasoning, coding | Google Workspace users |
Muse's unique advantage: it's the only agent that proactively surfaces reminders without being asked. ChatGPT/Claude/Gemini wait for a prompt. Muse notices your package ships tomorrow and tells you at 8 AM. This is the first time a major consumer AI has shipped proactive behavior at scale.
Muse's unique disadvantage: data access scope is the broadest of any consumer AI. It reads your actual messages in FB/IG/WhatsApp by default — not just metadata, not just "signals." ChatGPT/Claude process your prompts but don't read your other apps. If you have sensitive conversations on WhatsApp, Muse is not for you.
What can Muse actually do in 2026?
Muse's 12 confirmed capabilities (live as of Sept 22, 2026):
| Capability | Where it works | What you can ask |
|---|---|---|
| Cross-app memory | FB, IG, WhatsApp, Meta AI | "What did Sarah say about the trip?" |
| Proactive reminders | All 4 apps | Auto-reminders for packages, birthdays, flights |
| Multi-step booking | Meta AI chat | "Plan a Tokyo trip Oct 15-22 under $2K" |
| Voice notes (WhatsApp) | Dictate reminder while driving | |
| Reels search | "Show me hiking boots under $100" | |
| Marketplace search | "Find a used Peloton under $500 nearby" | |
| Messenger event detection | FB Messenger | "When is Sarah's birthday party?" |
| Multi-language translation | All 4 apps | Translate 100+ languages in chat |
| Image generation (Meta AI) | Meta AI | Generate images with Llama 4 / Spark 1.3 |
| Calendar extraction | All 4 apps | Auto-extract events from chat |
| Smart replies | Messenger, IG DMs | Context-aware reply suggestions |
| Cross-platform search | Meta AI | "Find that restaurant Sarah mentioned last week" |
What Muse can't do yet (Sept 2026): no third-party app integrations (no Notion, no Slack, no Gmail). no video understanding (audio + images only). no real-time web search (uses Bing via Meta AI).
How to opt out of Muse (or delete your data)
You can opt out of Muse entirely, or selectively disable per-app access. Step-by-step:
| Action | Path | What it does |
|---|---|---|
| Full opt-out | Settings → Privacy → Muse → Disable | Removes Muse from all 4 apps; deletes cross-app memory |
| Disable per app | Settings → Apps → [FB/IG/WhatsApp] → Muse toggle | Disable Muse in specific apps only |
| Delete memory | Settings → Privacy → Muse → Clear Memory | Wipes cross-app conversation history |
| Auto-delete after N days | Settings → Privacy → Muse → Auto-Delete | Choose 30/90/365 days |
| Disable voice notes | WhatsApp → Settings → Muse → Voice | Stops Muse from transcribing voice notes |
| Disable proactive reminders | Settings → Privacy → Muse → Proactive → Off | Keep reactive chat, disable reminders |
| Disable booking | Settings → Privacy → Muse → Booking → Off | Disables Booking.com, Airbnb, OpenTable, Expedia integrations |
Important: Opting out of Muse does NOT delete data Meta already collected via other means (ad tracking, profile data, message content for safety scanning). For full data deletion, use Meta's Download Your Information tool or delete your Meta account.
What's the difference between Muse, MUSE AI, and Meta AI?
Three different products that users confuse. Muse is the personal agent (Sept 8, 2026). MUSE AI (third-party) is an unrelated creative writing tool. Meta AI is the chatbot underlying Muse.
| Product | Owner | What it is | When launched |
|---|---|---|---|
| Muse personal agent | Meta | Cross-app AI agent in FB / IG / WhatsApp / Meta AI | Sept 8, 2026 |
| Meta AI | Meta | Standalone chatbot (reactive, like ChatGPT) | Sept 2023, rebranded 2024 |
| Muse Spark 1.3 | Meta | Underlying LLM powering both Muse and Meta AI | Sept 2, 2026 |
| MUSE AI (third-party) | MUSE AI Inc. | Creative writing assistant (unrelated) | 2022 |
| Muse (Apple, music) | Apple | Apple's music creation app for iPad (unrelated) | 2024 |
| Watermelon | Meta | Next-gen flagship model (Muse's successor) | Q2 2027 (projected) |
Most common confusion: "MUSE AI" (a small creative writing tool) vs "Meta Muse" (the consumer agent). They share the name but are unrelated companies. When searching, look for "Meta Muse personal agent" to find the right product.
Who is Alexandr Wang and why is he leading Muse?
Alexandr Wang is Meta's Chief AI Officer (CAIO), appointed June 2026 — the youngest CAIO in Big Tech at 28. He previously co-founded Scale AI (data-labeling company) at 19, sold it to Meta in 2025 for ~$14B, and joined Meta as part of the deal.
Wang is the public face of Meta's "personal superintelligence" strategy — a counter-narrative to OpenAI's "AGI for everyone" and Anthropic's "helpful, harmless, honest." Wang's framing: "superintelligence should be personal — your agent, your data, your life — not centralized."
Wang's role at Meta:
- Leads Meta Superintelligence Labs (MSL) — formed June 2026
- Reports directly to Mark Zuckerberg
- Oversees Muse, Spark, Watermelon (next-gen model), and Llama 5 (open-source)
- 2027 roadmap: ship a "personal superintelligence" agent that outperforms ChatGPT + Claude combined on agentic benchmarks
Why it matters: Muse is the first consumer product from Wang's MSL division. The bet is that consumer AI wins on proactive + cross-app + personal data — not just chatbot intelligence. If Muse succeeds, Meta captures the consumer AI surface that Google Assistant / Siri have failed to own. If it fails, Meta falls further behind in the AGI race.
For the broader October 2026 AI launch wave (Kolibri, Gemini 4 Argon rollout, Anthropic Claude for Government GA, Cloudflare Clef), see our October 2026 AI launches recap.
Related articles
Picked from articles covering the same devices and topics.
Meta Muse AI: 3 models in 3 weeks — Spark 1.3 (Sept 2), Glimmer (Aug), Muse agent (Sept 8)
Meta Muse AI: 3 products in Aug-Sept 2026 — Spark 1.3 (Sept 2, beats GPT-5.6 Sol on coding), Glimmer (Aug, open-weight), and Muse agent (Sept 8, FB/IG/WhatsApp). Watermelon in dev.
WoW Forever server status LIVE Oct 7: stable queues 12-35 min (post-Phase 2 stress)
WoW Forever server status Oct 7: ONLINE all 8 regions after Phase 2 stress test. Queues 12-35 min. 100K+ beta, full launch Oct 24. Live status + invite waves.
iPhone Duo specs: 5.4" cover + 7.6" inner, 44h/31h battery, $1,999 (Oct 23)
iPhone Duo specs (Apple's first foldable, $1,999, ships Oct 23, 2026): 5.4" cover + 7.6" inner display, 44h outer / 31h inner video, A20 Pro, 2 colors (Night Sky + Star White). Touch ID, IP68.