cd /news/ai-agents/ahead-of-meta-connect-muse-s-mac-app… · home topics ai-agents article
[ARTICLE · art-137888] src=runtimewire.com ↗ pub= topic=ai-agents verified=true sentiment=· neutral

Ahead of Meta Connect, Muse's Mac app shows an email address for your AI assistant

A static inspection of the Muse 2.2 Mac installer (build 1074644564, SHA-256 3db5892d1a2eed700554083edfa0c09daf1c8872ffd919c8117725a3722eec49) found a flag-gated client flow for creating an agent-owned @muse.ai email mailbox, RuntimeWire reported on September 23rd ahead of Meta Connect 2026. The Mail route is controlled by a remotely set flag named hatch_web:muse_mail that defaults to false, and the inspection did not establish that the service is live or that Meta plans to announce it at Connect. The CreateMailboxSection accepts handles of three to 32 lowercase letters, numbers or hyphens and proposes the sender name "Muse Personal Assistant (user name)"; Meta was not contacted before publication.

by read5 min views2 publishedSep 23, 2026
Ahead of Meta Connect, Muse's Mac app shows an email address for your AI assistant
Image: Runtimewire (auto-discovered)

A gated setup flow lets users choose an @muse.ai address for their agent, but the client inspection cannot show whether Meta will enable it at Connect.

        By [Ryan Merket](https://runtimewire.com/author/ryan-merket)
        · Published 
        · Updated 

RUNTIMEWIRE INVESTIGATION — Original analysis

Original reporting by RuntimeWire, based on reverse engineering.

Why it matters #

An @muse.ai mailbox would let people address a Muse agent directly. The Mac client shows the intended flow, while live availability and the controls on incoming mail remain untested.

Reporting record #

Finding

A static inspection of Muse 2.2 found a flag-gated client flow for creating an agent-owned @muse.ai mailbox, but did not establish that the service is live.

How we verified

Methods: reverse engineering.

The supplied Muse-2.2.dmg (SHA-256: 3db5892d1a2eed700554083edfa0c09daf1c8872ffd919c8117725a3722eec49) contains a Mail feature flag, CreateMailboxSection, @muse.ai address suffix, and mailbox and message requests in Muse.app/Contents/Resources/hatch/index.html.

RuntimeWire statically inspected the user-supplied Mac installer, checked its Info.plist for the version and build, and examined the packaged client code for the mailbox setup flow and requests. It did not test an authenticated account or live server.

Tested versions: Muse 2.2, build 1074644564 (production channel; statically inspected).

Reproduction

Reproduction does not apply to this reporting (document-driven).

Company response

The company was not contacted before publication.

Ahead of Meta Connect 2026, which begins later today (it's midnight here), Meta's Muse Mac app shows how users could give their AI assistant its own @muse.ai email address. A static inspection of a Muse 2.2 installer supplied for review found a gated address-creation flow and a mailbox with Received and Sent views. It did not establish that the service is live or that Meta plans to announce it at Connect.

The distinction matters to Muse's stated design: an agent that keeps working after a user closes the app and follows up when a task needs attention. A dedicated address would give that kind of assistant a direct channel for correspondence, instead of requiring every exchange to pass through its owner's existing inbox.

The inspection, conducted on September 23rd, examined a user-supplied Muse-2.2.dmg. Its Info.plist identifies version 2.2, build 1074644564, on the production channel. The findings come from client code packaged inside the Mac app. They do not establish that Meta has enabled the feature for users or that the mail service can deliver messages.

What the Mac client shows

In Muse.app/Contents/Resources/hatch/index.html, the Mail route is controlled by a remotely set flag named hatch_web:muse_mail. The client defaults that flag to false. When it is off, the route is removed from navigation and returns a not-found state if opened directly. The installer does not reveal the flag's live server-side value.

A CreateMailboxSection tells the user that the assistant does not yet have an email address and asks them to choose a handle. The address ends in @muse.ai. The client accepts handles of three to 32 lowercase letters, numbers or hyphens, and has a specific message for a server response indicating that a requested handle is taken. It proposes a sender name in the form Muse Personal Assistant (user name) and tells the user that recipients will see that name on mail sent by the assistant.

The code also defines requests to retrieve and create a mailbox, list messages and download attachments. Its message handling distinguishes incoming from outgoing mail and includes fields for sender and recipient, delivery status, sender-authentication results and warnings. The interface provides Received and Sent filters, attachment downloads, and warnings for unverified senders or likely spam. Those details support a narrower conclusion than a product launch: Meta has built a client for a dedicated agent mailbox, but the inspection did not test an authenticated account or a live server.

TestingCatalog reported on September 18th that an unreleased Mail tab appeared in Muse's web interface. At the time, the visible tab could have been either a separate mailbox or a place to display messages from connected personal accounts. The address-creation flow in the Mac client resolves that question about the intended interface. It does not settle availability.

A new inbox, and a new entry point

Meta's September 8th launch announcement described connecting existing email accounts, with users choosing whether Muse can read or send through them. Meta said Muse asks for approval before sensitive actions such as sending an email. An address belonging to the assistant would create a different path: other people could potentially contact it directly, without writing first to the user's personal inbox. The client code does not establish who would be allowed to send to such an address or what approval a reply would require.

That difference reaches the security design Tarek Sheasha, a Meta Superintelligence Labs engineer, described at launch. Meta says its separate Sentinel system controls network activity and permissions, while other defenses label external content as untrusted and screen for prompt injection. Sheasha also described filters that keep one-time codes and password-reset links out of Muse's connected-email workflow. His account describes Muse's launch architecture; the Mac client inspection does not show how those controls would apply to messages arriving at an @muse.ai address.

Meta offers Muse for Mac alongside its mobile apps. Going into Connect, the mailbox code shows a product direction within that client, not a confirmed announcement, release date or operating mail service. The consequential step would be enabling an agent to receive correspondence addressed to it while preserving the owner's control over what the agent reads and sends.

── more in #ai-agents 4 stories · sorted by recency
── more on @meta 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/ahead-of-meta-connec…] indexed:0 read:5min 2026-09-23 ·