v1.6.0

Scam guard

InboxScout now looks at every new email for the shapes scams take, on your computer, with plain rules. No AI helper is needed and nothing is sent anywhere.

It catches a "bank" writing from a free mailbox, a company name on an address that is not the company's, a grandchild who suddenly needs gift cards or a wire, a prize that needs a fee first, a "renewal" with a number to call, blackmail, pressure to act within 24 hours, requests to confirm a password or account number, and links that hide where they go. Thirty-five well-known companies and offices (PayPal, Amazon, Apple, Microsoft, Google, the IRS, Social Security, Medicare, the big banks, the parcel carriers, the antivirus vendors, and more) are known with their real addresses, so mail that truly comes from them is left alone.

When something looks wrong:

• It is the first card on Today, in red, with the reasons in plain words and one line of advice ("Call that person yourself, on the number you already have").
• A likely scam is filed under promotions and can never appear as a chore under Needs you or Waiting on a reply.
• Read aloud says it first. The morning report and Send me my brief get a "Looks like a scam" section at the top.
• A trusted helper gets a heads-up of its own, naming the sender and subject only.

Nothing is deleted or moved. The card says so. Details: docs/SCAM-GUARD.md.

Stop these emails

In Inbox review, newsletters and promotions that offer an unsubscribe link now show a Stop these emails button. It opens the sender's own unsubscribe page in your browser (or a ready-to-send message in your mail app). InboxScout still never sends anything itself.

Full Changelog: v1.5.7...v1.6.0

v1.5.7

Smoother setup

It knows your email service. Type your address and InboxScout picks Gmail, Yahoo, iCloud, or Outlook on its own (AOL addresses take the Yahoo path). The "Where is your email?" list is still there if you need it.

Fix it for me reconnects without an AI helper. When Google, Yahoo, or Apple stops accepting an app password (this happens whenever you change your main password), pressing Fix it for me under Settings → Health opens the service's own page. Sign in there, press Create, and you are reconnected. The automatic check before each scan never opens a window by itself.

Email us for help. Settings → Health → Details now has a button that copies the diagnostics and opens a message to hello@inboxscout.ai with a "press Paste here" note. Nothing is sent until you press Send, and the diagnostics hold no email text and no passwords.

Fewer moments of wondering. While the very first check runs, Today says "Your first brief takes about a minute." On the welcome screen, See an example brief shows a made-up morning brief before you connect anything.

Full Changelog: v1.5.6...v1.5.7

v1.5.6

Connecting your email without a fight

Get it for me. On the Connect your email step (and Setup → Email accounts) there is now one button that needs no AI helper. Press it: Google's, Yahoo's, or Apple's own app-password page opens in an InboxScout window. You sign in there and press Create. InboxScout picks up the new app password on its own, checks it, and connects the account. Nothing to copy, nothing to type. There is a Stop button, a 15-minute limit, and the window only allows the email service's own pages.

Plain answers instead of Command failed. If you paste your everyday password where the app password goes, InboxScout says so before trying it, and offers Try it anyway. If the service refuses a password, you now read "Gmail did not accept that password. It needs an app password, not your everyday password." instead of the server's raw reply. No-internet and 2-Step Verification cases get their own sentences.

Fixed. The half-cut error add': Error: Command failed came from InboxScout trimming the wrong part of an internal message; every screen now uses one shared cleaner.

Still to come

The true one-click Sign in with Google / Microsoft buttons are already built and turn on once the owner registers the free sign-in apps (Setup → Sign-in setup, or docs/GOOGLE.md and docs/OUTLOOK.md).

Full Changelog: v1.5.5...v1.5.6

v1.5.5

Full Changelog: v1.5.4...v1.5.5

v1.5.4

Releases are now built on all three platforms and published by one job only when every installer is present.

Full Changelog: v1.5.2...v1.5.4

1.5.2

v1.5.2: privacy policy and terms pages, dependency security updates

- site/privacy.html and site/terms.html (Cooper Studios LLC), linked from the
  site footer, the README, and the in-app terms.
- nodemailer 7 -> 10; undici and uuid pinned to patched versions under the AI
  SDK and MSAL trees. Shipped-dependency audit: no high or critical advisories.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01S8W5LYgLKjchYYdkHXLHC6

1.5.1

v1.5.1

release: export signing secrets only when they exist (empty CSC_LINK …

1.5.0

v1.5.0: reads attachments and photos

- Attachments are fetched from IMAP, Gmail, and Outlook (10 MB per file,
  budgets per message and per sync, tiny inline images skipped) and read on
  the person's computer: PDF, Word, Excel, CSV, text, HTML. Photos and scans
  go to the AI helper when it can see images, otherwise a bundled offline OCR
  (tesseract.js) reads them. Scanned PDFs without a text layer are listed as
  such.
- What is found (amounts, dates, people, document type, a plain summary)
  feeds the classifier and the built-in engine, so bills and deadlines that
  only exist inside an attachment reach the brief, with "from the attached
  invoice.pdf" as the source.
- Inbox review lists each file with its summary and Open; Today rows carry a
  small paperclip; Ask answers "what was in the PDF from…"; bridge ops
  list_attachments / read_attachment / open_attachment; phone read-only.
- Storage under userData/attachments with a health check and cleanup after
  attachmentsKeepDays (90); settings readAttachments (on) and
  attachmentsKeepDays with plain-words rows; docs/ATTACHMENTS.md; README,
  site, roadmap, Hermes skill.
- Packaging: OCR engine, worker, and English data unpacked from asar.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01S8W5LYgLKjchYYdkHXLHC6

1.4.0

v1.4.0: trusted helpers, ask about your mail, quality & speed sweep

- Trusted helpers wired into the pipeline (digests and heads-ups after
  scheduled/catch-up runs; helper replies surfaced as "Notes from your
  helpers") and onto Today (Ask for help per item, the 7-day notice, the
  notes card). README promise: "only ever sends to you and the helpers you
  name."
- Ask about your mail mounted on Today; the brief cache is invalidated after
  every run.
- README, site, and roadmap updated for both features; version 1.4.0.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01S8W5LYgLKjchYYdkHXLHC6

1.3.0

v1.3.0: InboxScout on your phone (iPhone & Android), Linux release

- Phone companion: Setup → On your phone shows a QR code; the desktop serves a
  phone-sized web app on the home Wi-Fi (check email, what needs you, bills,
  this week, promises, people, past briefs) with its own token, an
  allow-listed read-mostly API, Add to Home Screen support, and "New code" to
  revoke. src/main/api/phone.ts, phoneApp.ts, views/Phone.tsx, tests,
  docs/PHONE.md; settings phoneAccess (off by default) and phonePort.
- README and site: phone feature and FAQ; roadmap v1.3 shipped and v1.4 plan
  (trusted helpers, conversation, quality & speed).
- Version 1.3.0; first release with Linux AppImage and .deb.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01S8W5LYgLKjchYYdkHXLHC6