AI email assistant review
PureBox.ai review
Best for: Individual Gmail users with high volumes of routine clutter who want batch review, visible reasoning, Gmail-native archive/trash actions, and a cautious path from manual approval to selective automation.
Our review
The full take on PureBox.ai
PureBox is an active, newly launched Gmail-only AI inbox cleaner that classifies mail into Attention, Archive, and Trash tracks and keeps cleanup review-first by default. It is best suited to individual Gmail users who want to reduce newsletters, promotions, receipts, and other routine clutter without replacing Gmail. It is not a general-purpose email-writing assistant, and its early adoption evidence, single-account limitation, and conflicting privacy language require caution before purchase.
Features and fit
What PureBox.ai does
Scans Gmail through Google OAuth; classifies messages into 21 descriptive labels and three action tracks; uses sender behavior and mailbox history for personalization; explains recommendations; applies approved label, archive, trash, and mark-read actions; supports sender rules, continuous and real-time classification on Pro, opt-in automatic cleanup, history, and guarded Remote MCP cleanup plans. The privacy policy says it never sends mail on a user's behalf.
Product type
Hosted Gmail-cleanup web service with an Android app and a live hosted Remote MCP server for compatible AI clients.
Email and platform support
Gmail only. A Product Hunt maker response says one Gmail account is supported per PureBox profile; multiple-account support was described as planned. No current Outlook, Microsoft 365, IMAP, or multi-provider support was verified.
Pricing
Plans and cost
Free sample at $0 for one manual scan of up to 1,000 emails. PureBox Pro is $4.99 per month billed $59.90 annually; the homepage also lists a $7.99 monthly option.
Free plan or trial
Free one-time sample scan, no credit card required. This is a limited proof-of-value scan rather than an ongoing free plan.
Privacy and security
What happens to your email data
PureBox requests the gmail.modify OAuth scope, says it encrypts data in transit and at rest, and says raw message bodies are not retained after classification. Its privacy policy permits selected or partial body processing, stores derived Gmail metadata and classifications, and permits Gmail-derived data, feedback, and session replays to improve PureBox-specific models. Consent-gated PostHog and Better Stack recordings may capture Gmail-derived subjects, senders, snippets, and classification results rendered in the app. The vendor says Gmail data is not used to train generalized, third-party, or foundation models. See the conflicts and cautions below before publishing a broader privacy claim.
Security certifications and data-handling commitments are attributed to the vendor unless the supporting audit artifact was independently reviewed.
Independent signal
What users are saying
The engine found no meaningful on-entity PureBox discussion in its Reddit or TikTok results; the surfaced items were generic AI/email content or explicit entity misses and should not inform a product verdict. The separate Product Hunt launch thread provides early questions and maker answers, but no verified customer review evidence.
No meaningful named-product social signal from the engine; weak early launch signal from Product Hunt.
2026-07-10 through 2026-08-09
PureBox launched July 26, 2026 and had 159 Product Hunt points and 155 followers when checked, but no formal reviews
Confirms a recent public launch but not established customer validation
Launch commenters focused on false positives, regulated inboxes, archive searchability, and multi-account support
Shows the trust and workflow questions early evaluators consider most important
A maker response says the product currently supports one Gmail account per profile; a commenter identified that as a day-to-day blocker
Establishes a concrete current limitation, but only from a launch-thread response
The Android app was updated July 24, 2026, showed 5+ downloads, and had no visible rating aggregate
Supports active maintenance while showing adoption evidence is still very early
A generic automated review page incorrectly says PureBox was founded in 2010+, has 100+ employees, and uses contact pricing
Demonstrates why low-quality directory pages must not be treated as evidence; the page was excluded from product facts
Limitations
What to know before choosing PureBox.ai
- Conflicting model-training language: The security page says PureBox never uses email to train AI models. The privacy policy permits Gmail-derived data, classifications, user feedback, activity logs, and session replays to train or improve models narrowly tailored to PureBox. Do not publish the broader "no training" claim until the vendor reconciles these statements.
- Body-processing nuance: The security page says classification uses sender, subject, and a short preview and that full bodies are never stored. The privacy policy allows selected or partial message-body processing and says raw bodies are not retained after classification. Publish only the narrower retention claim.
- Subprocessor inconsistency: The privacy policy identifies Microsoft Azure OpenAI Service as the third-party AI provider, while the separately maintained subprocessor page says OpenAI is the current AI classification subprocessor. Reconfirm the actual provider before publishing provider-specific claims.
- Session-replay exposure: Consent-gated PostHog and Better Stack recordings may capture Gmail-derived subjects, senders, snippets, and classification results rendered in the interface. This is material context for a privacy-sensitive directory entry.
- Early-stage operational risk: The service is an individual sole-proprietor project that expressly says support is best-effort and unstaffed. Product Hunt had no reviews and Google Play showed only 5+ downloads when checked.
- Regulated use is excluded: PureBox does not offer a HIPAA BAA and says regulated healthcare mailboxes are outside scope. Its terms also say it is not a records-retention, compliance, spam-security, or mission-critical system.
- Narrow category fit: PureBox organizes and cleans Gmail; it does not currently present itself as a reply-drafting, auto-sending, meeting-scheduling, or full email-client assistant. Place it in an AI inbox-cleaner subcategory rather than treating it as a direct substitute for every AI email assistant.
- Single-account limitation: One Gmail account per profile is a notable blocker for people managing personal and work inboxes together.
- SEO evidence pollution: At least one generic automated review page published demonstrably incorrect company age, size, pricing, and audience facts. Ignore third-party pages that do not show product-specific evidence.
Evidence
Sources behind this review
Product facts come from current first-party pages. Independent and community sources are used for context—not as proof of every product claim.
PureBox works inside Gmail, groups clutter, and waits for approval before actions
The product uses 21 labels plus Attention, Archive, and Trash tracks, with low-confidence mail kept in Attention
Free covers one scan of up to 1,000 emails; Pro costs $4.99/month billed $59.90/year and adds ongoing classification
The homepage also lists a $7.99 monthly Pro option
PureBox uses gmail.modify, does not request the broader mail.google.com scope, and says it never sends mail
The privacy policy permits Gmail-derived data and session replays to improve PureBox-specific user-facing models
The security page says full bodies are not stored and email is never used to train AI models
The current subprocessor page lists OpenAI, Azure, Supabase, Better Stack, PostHog, Stripe, Google, and MailerLite
PureBox excludes regulated healthcare use and does not offer a HIPAA BAA
The hosted MCP server is live for compatible clients and exposes guarded scan, suggestion, sender, history, and cleanup tools