Meta Muse
Meta
As of 9 Oct 2026, PAEval has compiled 30 capability records for Meta Muse from 11 source URLs; 24 come from the vendor's own documentation and none has yet been independently observed. Assistant Benchmark (updated 8 Oct 2026) gives it 9.3 on its 1–10 scale, the mean of the 8 of 15 dimensions it scored, ranking #1 of 12 scored products. In micro1 PersonalAgentBench it completed 2 of 10 workflows without a critical violation (20% trusted completion; 95% interval 6–51%), and 2 of 10 counting any completion. Scores from different evaluators use different scales and are not combined. PAEval does not publish its own score or rank for it.
2026-10-09 Consumer/Small Business Muse; fixed Free/Power/Maximum, region, cloud or Mac connection status; not equivalent to Muse Spark model, Muse Code or the meditation app of the same name.
- Independent0
- Vendor24
- Third-party report1
- Not established5
30 records · 11 source URLs · 1 external result(s)
Capability summary
Strongest status per domain; every underlying record is listed below. Availability does not establish task success.
Positioning & scope
For personal long-term goals and daily affairs; the small business version is a new skill and connector within the same Muse, and independent enterprise isolation products cannot be inferred from the name.
Regions, eligibility & access
The official update confirms the United States and Canada; it is still a limited test and the complete list of countries has not yet been verified.
Requires 18 years old or local adult age and Meta account; iOS, Android, web, Mac entrance.
The US App Store lists English plus 38 languages, including Simplified/Traditional Chinese; iPhone requires iOS 18+.
Limitations: The list of store localizations does not represent the quality of implementation in each language or the locale's registrability.
Pricing & quotas
Free has a limit; Power is US$20/month, 500M Muse tokens/week; Maximum is US$100/month, 3B/week; automatic monthly renewal.
Limitations: Rights are displayed by region and account; proprietary tokens are not converted into task numbers or other manufacturer tokens. The free quota value has not been verified.
Autonomy & background work
Continue after the app is closed; proactively identify emails to handle, make suggestions, and follow up based on ongoing goals.
The documentation is clear about cron/alerts/monitoring; some connectors will proactively update agents for calendar changes.
Limitations: The general webhook range, minimum frequency and scheduling upper limit have not been verified.
Execution environments
Independent Linux cloud VM per user, with browser, file system, terminal; executable code, sub-agent and cron.
The Mac version is authorized to process files, browser tabs, Messages, Calendar, Notes and other native applications.
Conditions: The user grants corresponding Mac permissions; not just remotely viewing the cloud desktop.
Independent Android/iOS cloud phone or native phone GUI automatic execution has not been confirmed. The Android SMS/Apple Health connector and mobile chat portal cannot be used as cloud mobile evidence.
Calls, messages & identity
Can read, draft, send, organize emails and schedule/reschedule calendars; important sending is approved by default, and read-only and write permissions can be distinguished.
Conditions: Connect to the relevant account and authorize writing.
On September 16th, the outbound call beta of American merchants was released through social posts by the person in charge, and 404 reports were used to make calls, make appointments, and return summaries/texts.
Limitations: This round was verified through media citations, and the original post was not directly read; internal testing may have manual assistance, and the proportion and current public coverage are unknown.
Unconfirmed, you can give Muse an independent email/phone number or cross-user team identity; sending on behalf of the user does not equal an independent identity.
Connectors & tools
The built-in API connector can be read and written, and some events can be actively pushed; it may be automatically connected to Facebook/Instagram/Threads in Accounts Center.
Conditions: The connector description and user authorization determine the actual reading and writing.
Asana, Box, Canva, Dropbox, Figma, Notion, Shopify, Slack, Stripe, Zoom and Meta business accounts have been announced; the catalog is dynamically updated within the app.
Custom connectors can be written for services with API/CLI; self-built connectors are not subject to Meta review.
This round has not verified the universal MCP access/API task interface of Muse consumer products; Muse Code MCP documents are not used to prove Muse personal agent support.
Memory & personalization
Long-term memory can be forgotten according to the theme requirements; its saved files and memories can be viewed/edited/downloaded, and the acquired information will not be automatically cleared when the connector is disconnected.
Resetting will delete chats, files, and active tasks; memory/file control and disconnecting services are different operations.
Channels & interfaces
Muse mobile/Mac/web and WhatsApp chat, you can close the application and continue to receive results.
Publicly available materials confirm that voice is an input/interaction data type, but this cycle does not confirm the current scope of universal real-time two-way calling, actively calling users, or joining meetings.
Research & deliverables
The official website lists browsing research, image generation and documents; the Library document lists documents, images, videos, podcasts and other products.
Limitations: Format fidelity, editable export and long task quality are not proven by official statements.
You can write code, compile and generate the tools required for tasks on the cloud terminal.
Transactions & bookings
Browse, fill out forms, shop and book; Link one-time card and qualifying purchase protection, payments subject to approval.
Limitations: Shop Pay and Confidential VM are still coming soon according to the announcement, so they cannot be counted as already online. Merchants may reject automation.
Collaboration & parallelism
Official description of multitasking and concurrent subagents; no concurrency hard cap announced.
Small Business Connector does not demonstrate multi-person shared agents, member roles, or private/team memory isolation; not verified for this round.
Reliability & recovery
The browser can be viewed and taken over, with action auditing; the VM is continuously backed up and can be restored, officials say.
Limitations: Recovery point objectives, snapshot history UI, task-level rollbacks, or SLAs are not verified.
Permissions, security & data
Sentinel independently controls connector actions and network exits. Authorization can be once/session/task/time limit/continuous; fine-grained read and write restrictions are based on services.
Limitations: It is an official architecture statement, not a zero-vulnerability guarantee; prompt injection is still a known risk.
Training is turned on by default and can be turned off, and the settings for backtracking apply to previous interactions; VM/chat is not handed over to the Meta advertising system, but external behaviors generated by visiting merchants may still affect advertising.
The current Secure VM does not mean that Meta cannot be read; the Confidential VM will be launched later/small-scale trusted testing in the official article.
External measurements & hands-on results
Different methods, samples and dates. Not combined into one success rate.
Single reviewer Muse business tasks, access rights are self-limited ↗
- Evidence type
- independent_small_sample_test
- Tested
- 2026-09-27
- Sample
- planned tasks: 10; scored tasks: 6; unscored tasks: 4
- Findings
- Reported errors in determining email completion status, weak web page products, and observed context-driven meeting preparation and completion studies after closing the app.
- Limitations
- The four unscored items include denial of permission, undecided, and not run, and cannot be recorded as four failures.; The subjective overall impression is separated from the scored task results; it is not integrated into the PAEval total score.; Single account, short window, no long-term stability/security audit.
Analyst notes
Key differences
- The same product covers life affairs and Meta small business connector, and is already implemented in Mac native applications.
- The consumer side provides free quota, and payment is based on weekly tokens, which cannot be directly compared with the task/credit package.
- You can request forget by topic; however, Meta connections, default training, and future Confidential VMs under the same account need to be marked separately.
- The external response is displayed as beta/media evidence for US merchants, and is listed separately for human assistance; do not label it as a mature global automated phone agent.
Known unknowns
- Full country/mission language list
- Universal real-time voice, user callback, conference joining
- Independent cloud phone
- Personal product MCP/API task interface
- Multi-person collaboration and isolation
- Concurrency/scheduling hard caps and task rollback SLAs
Source conflicts and how they were resolved
- topic: regions; older: First published in the United States on September 8; newer: September 29/October 8 Small Business in the United States and Canada; resolution: The latter indicates that the region has been confirmed; the full list is still unknown.; source ids: muse_launch; muse_smb
- topic: secure_vs_confidential; resolution: Secure VM has been provided; Confidential VM is only for preview/small-scale testing, and stronger privacy guarantees cannot be written into the current version in advance.; source ids: muse_security
- topic: phone_calls; resolution: The outbound call beta and internal manual assistance reports are divided into two pieces of evidence, and it does not assert that all public calls are made by humans.; source ids: muse_calls
Detailed capability observations
34 capabilities in 8 groups, reviewed cell by cell. Unknown cells were not researched or lacked sufficient public evidence.
| External contact and communication | ||
|---|---|---|
| User initiates real-time voiceTwo-way real-time voice between the user and his own agent; it does not mean that the agent calls out to a third party. | Unknown | Not researched yet |
| The agent proactively calls the userAgent initiates a call to the user; records in-app calls or phone network. | Unknown | Not researched yet |
| Outbound calls to third partiesDial business or personal numbers as authorized and have two-way communication; voice chat and generated audio are not… | Limited or conditional | The report quoted the person in charge announcing the beta of outbound calls to US merchants; the official original post was not read directly and cannot be regarded as proof of full openness or pure AI completion. |
| Answer external callsA third party can dial into a number or call portal controlled by the agent, and the agent will actually answer the… | Unknown | Not researched yet |
| Send and receive messages across channelsActual sending and receiving via email, SMS/RCS, WhatsApp, Slack, Teams, etc.; only drafting, not sending. | Documented | Official support is provided for drafting, replying and sending via authorized email addresses. |
| Participate in live meetingsActually join the meeting and listen, transcribe or speak; recording each operation and reading the meeting minutes is… | Unknown | Not researched yet |
| Agent independent external identityAgents have independent email addresses, numbers or platform identities; they are strictly distinguished from borrowed… | Unknown | Not researched yet |
| Execution environments and devices | ||
| Cloud browser operationThe agent opens web pages, operation forms and cross-site processes in the cloud; ordinary search interfaces are not… | Documented | The official provides an independent cloud browser that supports web page and form operations. |
| Cloud computers and terminalsAgents have operational desktops, file systems, and/or terminals; record persistence and isolation units. | Documented | Official description of persistently isolated Linux cloud computers. |
| Cloud Phone and Mobile AppThe agent can operate apps in cloud Android/iOS or equivalent mobile systems; mobile clients and mobile remote control… | Unknown | No evidence was found that the product natively provides a cloud phone that can operate the mobile system/App; the mobile client and cloud computer cannot be used as alternative evidence. |
| User computer operationAccess files or operate applications on authorized user computers; clarify system, online conditions and operating… | Documented | Muse for Mac can process native files, apps and browser tabs after authorization. |
| User mobile phone operationReading and writing the system/App or operating UI on the authorized user's mobile phone; only notification and… | Unknown | Not researched yet |
| Application connections and tools | ||
| Connector readRead accounts, mail, calendars, files, etc. through supported connectors; log services and read objects. | Documented | Officially listed are connectors for email, calendaring, and enterprise tools. |
| Connector writeCreate, modify, or send via a supported connector; having a read connector does not imply write access. | Documented | Official instructions can handle email sending and calendar changes through authorized accounts. |
| Custom connectors and toolsSupport users/developers to access custom APIs, MCPs or plug-ins; record protocols, hosting and permissions. | Documented | Custom connectors are officially supported; do not assume that all MCP deployment types will have access. |
| Login and session persistenceYou can securely authorize services and reuse the login for subsequent tasks; it is not equivalent to obtaining the… | Unknown | Not researched yet |
| Continuous execution and initiative | ||
| Offline background promotionAfter closing the client, authorized tasks can still continue; the duration and amount are separately noted. | Documented | Officials said that they will continue the commissioned work after closing the app. |
| Timing and recurring tasksRun tasks by date, time zone, or recurring schedule; only calendar reminders need to be noted. | Documented | Official instructions can set reminders and monitor regular messages. |
| event triggered executionIt is started by events such as emails, messages, Webhooks, and file changes; scheduled polling is a separate… | Unknown | Not researched yet |
| Proactive discovery and follow-upOpportunities, reminders or suggestions are found when no new instructions are received; independent writing requires… | Documented | Official descriptions will proactively provide suggestions and drafts. |
| Memory and personalization | ||
| Long-term memory and preferencesPreserve and use preferences, context, or project state across sessions; don't treat context windows as long-term… | Documented | The official description remembers user information and preferences. |
| Cross-channel continuationDifferent entries can continue the same task and related context; message mirroring and context can be separated. | Unknown | Not researched yet |
| Memory review and correctionUsers can view, correct or selectively forget; distinguish between full account reset and selective deletion. | Documented | Official support requires agents to forget specific content; other memory management operations are subject to verification separately. |
| Achievements and realistic tasks | ||
| Research and traceabilitySearch, synthesize and present verifiable sources; source accuracy documented by external review. | Unknown | Not researched yet |
| Documentation and multimodal resultsCreate or modify available products such as documents, tables, presentations, images/audio and videos; format as facet. | Documented | Official documentation and image generation are listed. |
| Coding and runnable deliveryImplement, run, test code, and deliver the application or change; only generating snippets is distinguished from… | Unknown | Not researched yet |
| Shopping reservation and service processingActual execution of shopping, reservation, unsubscription or service process; stopping at draft, shopping cart,… | Documented | Officially supports authorized shopping, reservations and Link checkout. |
| Collaboration and task closed loop | ||
| Cross-application multi-step processThe same task transfers information and executes across multiple applications; multiple independent plug-ins are not… | Documented | Official description of the multi-step web and connector workflow. |
| Parallel tasks and agent collaborationMultiple tasks/agents advance and hand off simultaneously; logging whether files, sessions, and identities are shared. | Unknown | Not researched yet |
| Failure recovery and waiting for follow-upIt can be recovered after failure, waiting for reply or external changes; the completed effect must be supported by… | Unknown | Not researched yet |
| Authorization, security and auditability | ||
| Fine-grained authorizationAllow, deny, or request confirmation by action, object, and scope; presence of controls does not equal security score. | Documented | The official provides permission options such as single permission, continuous permission and denial. |
| Manual takeover and continuationIt can be handed over to the user to complete the sensitive/blocked steps, and then returned to the agent to continue. | Documented | Officially allows users to take over shared browsers. |
| Activity and result trackingDisplay tool actions, execution records, receipts, task status or sources; the display range is noted separately. | Documented | Official notes show audit records of executed and planned actions. |
| Export, Undo and DeleteSupports data export, connection cancellation or task/data deletion; the three types of operations retain evidence… | Unknown | Not researched yet |