Should you connect your real mailbox?
Start with SuperLocal's fictional mailboxes if your goal is to evaluate its SDK architecture. Keep an existing mail client available until you have reproduced the test failures and checked the provider operations you need. Our evidence supports an evaluation checkout, not a recommendation to move essential mail into it.
The README lists Gmail and Inbound.new as implemented, while IMAP/iCloud remains in progress. Restricted Google login gives each approved person private mailboxes and settings; it does not provide shared-mailbox or team roles. These distinctions matter more to an adoption decision than the presence of a unified inbox alone.
What our sandbox actually established
Our run took place on September 28, 2026, at commit d3725bc. The retained build output shows an Inbox SDK TypeScript build, and that command exited zero. It does not establish that the browser app started, Docker deployed, OAuth completed, or mail was synchronized, sent or recovered.
The test command exited 1 after 37 seconds. The attached stored parser summaries record 338 passed, 337 failed and 675 total tests, with zero skipped. Treat those as recorded parser summaries, not a verified count of every unique test in the repository. The raw footer reports its own 31.02-second duration; the harness records 37 seconds for the command step. Those are different recorded timings, so neither replaces the other.
The retained tail names failures in authenticated SVG media handling, bounded thread ordering, private AI configuration and AI triage service behavior. It contains substantially more than three failure entries. One concerns archive inventory accounting and includes the word "skipped" in its test name; it is a failed test entry, not evidence of skipped tests. The tail does not establish a common cause or demonstrate an exploitable security flaw.
Our harness records command success, exit codes, elapsed time, retained output and parsed test counts. The matching stored record is the evidence for this run. The MrKeyoor testing methodology explains that reporting approach and its limits; its current description does not establish which methodology version applied in September. This correction adds no execution. The record contains no audit result, and scan coverage is unknown.
What the fictional demo lets you evaluate
The commit-pinned README documents two fictional mailboxes running through the real Inbox SDK without provider credentials. That gives a developer a starting point for examining the client and storage model before authorizing access to personal mail. We did not verify those interactions in a browser.
The SDK normalizes mail into SQLite while keeping provider identities, credentials and capabilities separate. Done and snooze are local workflows. The README also describes isolated HTML rendering and authenticated media routes. Those are implementation descriptions, not a security assessment; the recorded media-related test failures still need investigation.
Gmail recovery and aliases need specific checks
The supplied Gmail reconnect pull request identifies d3725bc as its before commit. Its author reports that an existing Gmail connection in reconnect_required could not recover through the offered sign-in action, and proposes wiring the existing reconnect flow. The supplied page shows an open pull request. Its candidate-build checks are contributor evidence, not our sandbox results or proof that the fix is merged.
The alias issue also needs a narrower reading than "aliases are unsupported." The maintainer says compose alias selection already works and reply selection mostly works when the alias appears in To/Cc. The reported gaps concern Bcc or list mail, multiple matching aliases and displaying the receiving alias on the inbox row. If these cases matter to you, include them in your own qualification rather than assuming the general sender-identity description covers them.
Persistent state is part of setup
The README instructs Docker operators to retain the volume mounted at /persist and back up configuration, databases and keys together with the app stopped. Missing or mismatched keys fail closed. Browser recovery copies and preferences also remain in the browser, so a server-volume backup is not a claim that every piece of client state is preserved.
Remote installations require Google OAuth configuration, an exact email allowlist and HTTPS. TLS and reverse-proxy setup remain external. Google application login does not itself connect Gmail or grant mailbox permissions. Before exposing an installation, verify both application access and mailbox authorization separately.
The stored license signal is null. That does not establish reuse permission or independently prove that every possible license location was checked. Confirm the applicable terms before incorporating the SDK into another product.
Which alternative fits your job?
Mailspring provides compiled desktop clients for Windows, macOS and Linux. Consider it when you want to install a desktop mail client rather than evaluate SuperLocal's host and SDK architecture. Its README describes a local sync engine and GPLv3 source; we did not test its compatibility with your accounts.
Roundcube is a browser-based IMAP client. Consider it when IMAP webmail is the requirement. It is a replacement choice for that client role, not a connector that completes SuperLocal's unfinished IMAP support. Its repository warns that the Git snapshot is not a stable version, so follow its release and installation guidance rather than treating the default checkout as a production release.
How to read the scores
These are editorial assessments, not measured product quality. The scale is 1 for a major obstacle, 2 for substantial unresolved work, 3 for usable evidence with material gaps, 4 for detailed supporting evidence, and 5 for strong evidence across the assessed scope. Setup considers configuration and retained state; documentation considers the supplied operational guidance; community considers visible maintainer and contributor interaction; maturity considers the recorded command outcomes and unfinished capabilities. Community activity does not establish adoption or support reliability.
