Loopky v1.7.0

Added

• Loopky connector for Claude (mcp): A hosted, multi-user MCP server, so Loopky decks can be created and edited straight from a Claude chat with no CLI install. Users sign in with Pubky Ring over OAuth 2.1. Each user gets an isolated repository graph, credentials are held in an encrypted vault, writes are rate-limited per user, and the server sits behind an edge flood guard. Every tool has a title and correct destructive hints.
• plugin: Route agents to the CLI or the connector by environment, putting the connector first in a sandbox
• plugin: Load the skill when a loopky command fails
• cli: Exit 17, request_refused, for an indexer 4xx
• auth: Sign an un-backed-up held key back in when its session is rejected

Fixed

• mcp: Many hardening fixes before launch: token and code revocation on unlink and on refresh reuse, session binding per principal, bounded sign-in slots, and a health check and watchdog that restart a wedged server
• cli: Login refuses a session from the other network, and its exit 8 carries its result and names what it replaced
• cli: Limit request_refused to request-shape statuses
• cli: Send deck edit's cover-url warning through onNote
• import: Detect spaced pipe and semicolon pairs, and keep them from outranking the real separator
• nexus: Page resources by timeline only, falling back only when taggers_count is refused
• tags: Fold language tags where they are derived
• srs: Retry a failed chunk read, read a chunk before writing it, and refuse a grade on unread review state
• home: Show counts as unknown when review state did not load whole; only the affected deck loses its badge
• auth: Sign in with a held key through grant on Android (#130), and lift the cookie pin now that Ring v2.0 parses grant URLs (#321)
• ios: Sign up with a cookie session, as Android does
• ios: Fetch Nexus and remote images off URLSession.shared, keep images on reappear, and finish each background task exactly once
• pubky: Count a client id's length in UTF-8 bytes, as pubky does

Full changelog: v1.6.0...v1.7.0

Loopky v1.6.0

Added

• Loopky Recap (Android and iOS): from 16 December, a yearly recap of what you studied, with share images for each network and a Home banner. Its Pubky post carries the loopky-recap label.
• Study options on a followed deck: Edit on a deck you follow now opens your own study options (Listen, Speak, Type the answer, Both directions). They are stored on your subscription, so the author's deck is untouched and you keep their updates. Making a copy is one step further in.
• Password-manager backup: save the recovery phrase to your password manager and restore from it, on Android and iOS.
• Tag folding: every tag write is folded to one canonical spelling, so café/cafe and foo bar/foo_bar are the same tag. Search folds the query the same way.
• Study defaults: the new-card goal defaults to 10 a day (was 20), and a try-mode sample is 5 cards (was 10).
• iOS: universal links that arrive as a browsing activity now open.
• CLI:
  • deck create, card add and import report card-writing findings as card_advice, and refuse --listen/--speak without a language pair.
  • A per-deck file lock stops two commands writing one deck at once.
  • card check-images checks the pictures on a stored deck, and paces itself per host on a 429.
  • Long writes report progress, also under --json.
  • --tag is folded and reported as tags_normalized.
  • Deck descriptions over 500 characters are refused.

Fixed

• Following a deck again keeps your study options.
• Backup: a rotation while the credential sheet is up no longer locks the screen, and a dismissed check after a save is no longer reported as "nothing was saved".
• Android: the signup and restore forms scroll clear of the keyboard; screens stay clear of the cutout and a side navigation bar in landscape; the code scanner no longer forces portrait.
• Tags count length in code points, and emoji survive the fold.
• CLI: card edit can clear a card's image, a resumed import is locked, findings name the right file line, and blank languages are refused.

Full changelog: v1.5.0...v1.6.0

Loopky v1.5.0

Added

• cli: loopky card reorder puts a whole deck in a file's order, and loopky card mv moves one card in the study order. A reorder that is interrupted can be resumed; until it is finished the deck refuses card writes with exit 9 and says how to finish it.
• plugin: The agent skill teaches card reorder and card mv, and what building a whole show's decks taught (#438)
• android, ios: Prompt for a newer version, at most once per local day — Play's flexible in-app update on Android, an App Store prompt on iOS
• backup: A once-a-day backup reminder, shown as a bottom sheet on Android and a sheet on iOS
• home: Lead the Today card with the daily goal
• ios: Recall the front picture on the card back as a miniature
• ios: Use Android's card text sizes on an iPad

Fixed

• cli: card mv refuses a self-anchor before reading, and never reports a stale ord
• plugin: Say that card add re-adds a card once it has a picture
• ios: Lay the phrase quiz options out in two columns
• ios: Pin the prompt label and miniature to the top of the card back
• ios: Let the study card fill the height on a tablet, and raise its picture cap there
• android: Drop the study card's height ceiling on a tablet

Full changelog: v1.4.0...v1.5.0

Loopky v1.4.0

Added

• Follow from a preview (Android, iOS): finishing a signed-in preview now offers Follow as the primary action, a guest who signs in mid-preview follows the deck they were looking at, and a follow made there offers to announce it to your followers.
• AI import prompt box (Android, iOS): the import screen's AI button is replaced by a prompt box, built in one shared place, that points your agent at the Loopky plugin first.
• cli: loopky doctor and the allowlist docs recommend the subtitle hosts and the new picture-lookup hosts.
• plugin: Teach the skill to build a language deck from a TV series
• plugin: Teach the skill more picture sources, in a fallback order
• plugin: Offer to report session friction as a Loopky issue, asked only at the end, before gathering anything

Fixed

• plugin: Keep only the subtitle sources a sandboxed agent can download from

Full changelog: v1.3.1...v1.4.0

Loopky v1.3.1

Fixed

• ios: Collapse long deck descriptions behind Read more
• ios: Show card thumbnails in the deck detail card list
• plugin: Ship no installer; loopky is a prerequisite the user installs
• ios: Drop ITSAppUsesNonExemptEncryption so uploads are accepted

Full changelog: v1.3.0...v1.3.1

Loopky v1.3.0

Added

• ios: Open loopky.app and pubky:// links in the app

Fixed

• ios: Stop background tasks rescheduling forever, and after a 507

Full changelog: v1.2.4...v1.3.0

Loopky v1.2.4

No user-facing changes — maintenance, tooling and documentation only.

Full changelog: v1.2.3...v1.2.4

Loopky v1.2.3

Changed

• plugin: The loopky skill installs the CLI with installer scripts shipped in the plugin, which check the release's SHA-256, instead of piping a download into a shell. Run /plugin update loopky to get it.
• store: The Play listing is shorter and plainer, titled Loopky Flashcards.

Full changelog: v1.2.2...v1.2.3

Loopky v1.2.2

Changed

• plugin: The Claude Code plugin no longer runs an install hook when a session starts. The loopky skill installs the CLI when it's missing, as before. Run /plugin update loopky to get it. In a cloud environment, you can put the install line in the environment's setup script instead:

  curl -fsSL https://github.com/jvsena42/loopky/releases/latest/download/install.sh | sh

Full changelog: v1.2.1...v1.2.2

Loopky v1.2.1

Fixed

• plugin: The loopky skill for Claude Code and Codex splits a card with alternative answers (vegetariano / vegetariana) into two, and fixes a card that breaks a rule before showing you the list, instead of noting the problem underneath. Run /plugin update loopky (Claude Code) or codex plugin marketplace upgrade loopky (Codex) to get it.

Full changelog: v1.2.0...v1.2.1