Changelog

Users can now convert ChatGPT data exports into Markdown, PDF, or plain text directly in their browser without uploading their history.

• Download selected conversations as Markdown, PDF, clean JSON, or plain text.
• Search and select specific conversations to export locally.

Public pages now display a custom image when shared on social platforms like X and LinkedIn, replacing the previous blank placeholder.

• The card includes the Keep logo and the page title.
• Developers can customize the card title for specific pages.

Learn how AI agents build and use context with an interactive guide featuring seven demos.

• Play with interactive demos to understand how AI agents build and use context.
• Visualize concepts like context windows, the agent loop, and context rot.

Reddit items now generate full summaries when enabled, and items are correctly identified as having context based on descriptions and notes rather than organization tags.

• Reddit feeds now generate full summaries and content.

New users receive a welcome email with onboarding resources and links. The browser extension is now easier to discover in the app menu.

Users can learn Markdown through interactive examples and printable quick references.

• Live syntax demos for headings, lists, and code blocks.
• Interactive playgrounds for tables and Mermaid diagrams.

A new tool lets you upload PDFs and convert them into Markdown format.

• You can copy the result or download it directly.

The token counter now handles special token strings correctly and can compare URLs and Markdown. Public tools now use a consistent layout and spacing.

• Token counter now correctly counts special token strings found in LLM documentation.

Content now displays correctly on small screens, and data is no longer shared between accounts.

• Content fits better on mobile screens.
• Prevents content from appearing in the wrong account.

The signup process now shows the verification step right away and handles email resends correctly.

You can now add content to notes without specifying the current revision. The system automatically handles conflict resolution.

• Replacements still require a revision to prevent accidental overwrites.
• Appends are now safer against concurrent edits.

Notes now load more efficiently by showing summaries or specific sections first, with the option to view the full content when needed.

• Jump directly to specific lines within a Note without loading the entire document.
• View a summary of Note revisions before deciding to read the full text.

Users can now specify distinct login commands for Claude Code and Claude Desktop to ensure correct attribution and access.

• Run keep login claude-code for Claude Code.
• Run keep login claude-desktop for Claude Desktop.

Source credentials are now redacted by default to protect your secrets. Note updates now preserve existing tags and projects unless you explicitly clear them.

• Explicit clear flags prevent accidental deletion of tags, projects, and state.
• Write receipts show exactly which fields changed.

You can now see the save status and save options at the bottom of the screen while scrolling through long notes.

You can now manually trigger a save and see your name or You instead of your email in revision history.

• See a live countdown before your notes autosave.

Shared links to Keep now automatically save and open the tagging interface. Legacy shared links also work reliably.

Users can now sign in to the Keep CLI with a browser instead of creating an API key. This provides a named, revocable credential for connected tools.

The account page now clearly shows your current active plan and allows you to switch between monthly and annual billing. You can also manage X bookmark sync settings directly from the account page.

• Toggle X bookmark sync on or off.

Saving links on mobile is now smoother and the app recovers from connection issues instead of freezing.

• Mobile users can save links without unexpected keyboard popups or scrolling issues.
• The app now recovers from connection problems instead of freezing or showing errors.

Users can now choose between a free Starter plan with limited notes or a paid Keep Pro subscription with advanced features.

• Keep Pro offers monthly or annual billing options.
• Optional X Sync is available for Pro users.

Capture returns more quickly without waiting for AI processing to complete.

You can now save links and paste content immediately. The saved items stay accessible in your Feed while background processing finishes.

• Items remain available even if processing is paused or hits limits.
• Failed saves are now shown as processing instead of errors.

Keep now supports session integrations for Pi and OpenCode, along with a Cursor preview. Documentation has been updated to clarify which agents are supported.

• Use Pi and OpenCode directly within Keep sessions.
• View Cursor as a preview option.

You can now rename, edit details, and delete tags directly from their detail pages. This makes it easier to keep your tags organized.

• Use the action menu to update tag names or descriptions without leaving the page.

The CLI is now available as keep@2.0.0 with preserved settings. Users can now use a canonical search command and integrate AI tools like Codex and Claude Code with project-specific context.

• Existing configuration and credentials are retained during the update.
• AI tools can now access project-specific notes for better context.

The app is now fully responsive on mobile devices, allowing you to navigate and capture items more easily.

• Capture items faster by setting collection, tags, and notes before saving.
• Use mobile-friendly sheets and controls for a smoother experience.

The app loads lighter and handles large feeds more efficiently.

• Scrolling through large lists of notes is now smoother.
• The app starts up faster with a smaller initial download.

You can now manage your account details, view invoices, and configure API keys directly in the app.

• View your plan usage and invoice history.

Users can now configure RSS, YouTube, X, and Hacker News sources with backfill and rollup options. Email inbox confirmations and Kindle address management are also restored.

• Manage email inbox confirmations and Kindle addresses.

You can now archive notes to clean up your workspace. You can undo an archive within ten seconds if you change your mind.

• Prevents archiving notes with unsaved changes.
• Restores the note content when undoing an archive.

You can now import from Pocket, Instapaper, and other services. You can also export notes as ZIP files and drag and drop Markdown files to create notes.

Markdown lists now render with proper bullets, and inactive formatting markers stay out of the way until you need them. External links in feeds show their labels by default and reveal their source when selected.

• Images from GitHub repositories now load correctly.
• Failed remote images are hidden cleanly instead of showing broken placeholders.

Background RSS sources now work with legacy Reddit saved-feed URLs. YouTube sources are more reliable by updating channel information and retrying temporary errors.

Search results appear more quickly, and you can now filter by tags and collections.

The CLI can now automatically generate summaries of AI sessions when they end. You can opt in to save detailed notes about your work and reasoning.

• Summaries capture decisions, files, and URLs.
• Raw transcripts stay local to the AI client.

Keep now detects your Git project and path when creating notes, making it easier to retrieve context later. The CLI also supports adding tags and custom properties to notes.

• Notes created inside a Git repository are automatically tagged with the project name and file path.
• The CLI now allows you to add custom tags and properties to notes for better organization.

You can now create named, revocable credentials for AI tools and CLI clients. Manage access independently and revoke specific clients without affecting your personal key.

• Use keep auth add to create a trusted identity for each tool.
• Rotate or revoke client credentials individually.

Users can now download a ZIP file containing all their current and archived notes as Markdown files. This feature enables local backups and data portability.

• The CLI now supports exporting the entire note library.
• Exported files preserve note properties and archive status.

Notes open in a dedicated reader page and items have stable permalinks that preserve your feed filters.

Users can now interact with their personal context library through a dedicated AI skill and CLI tools. This includes improved search and sharing capabilities.

• An updated Keep skill replaces the previous markdown skill for AI agents.
• The CLI now supports installation, status checks, and updates and the new notes format.

Users can now create and manage personal notes with search and archive capabilities.

• Notes support Markdown formatting.

Fixes an issue where YouTube videos sometimes display a generic placeholder title instead of the actual video title. Users will now see the correct title for their saved videos.

Fixes errors that prevented RSS subscriptions from working correctly in the extension.

• Subscriptions now work reliably with RSS feeds.

Extension saves now start fresh with tags and collections. Draft content like the current tab and URL is preserved.

• Extension saves no longer carry over tags or collections from previous saves.
• Draft content is retained for the new item.

You can now edit the title, tags, collection, and notes for items you have already saved. This allows you to correct information without visiting the web app.

• Update tags and collections for already-kept pages.

View Twitter replies and export them as Markdown. This lets you save conversations for offline reading or archiving.

• See direct and nested replies.
• Download conversations as Markdown or JSON files.

The popup opens instantly and captures page metadata like author and description automatically.

• See a live preview of your saved note.
• Last used collections and tags are remembered per site.

You can now save items with title, notes, and tags in a single command. You can also re-extract content using personal API tokens.

Users can now choose whether to generate AI summaries for new items. This helps reduce noise in feeds and saves resources.

• Users can disable summaries for specific sources.
• Users can turn off summaries globally to save resources.

Pasting Reddit links now retrieves the actual post content and comments instead of a verification page.

• Share links and deleted posts are handled correctly.

Mobile users can now navigate back to the feed from any settings page.

• A floating menu button appears on mobile to show core navigation options.
• This prevents getting stuck in nested settings pages.

Free plan saves now display the actual page title instead of the raw URL, making your saved links easier to identify and read.

Users now see upgrade options when they reach their content limits and can easily log out from the sidebar. The app also displays when usage limits reset.

• Upgrade prompts appear when users exhaust their content credits.

Users can now choose to save link-only items from RSS, YouTube, and X Article sources without spending credits on full content extraction.

• Link-only items are saved without extra page extraction.

Free users now have a consistent way to view upgrade options and understand which features are available with a paid plan.

• Upgrade options are now easier to find in the feed and settings sidebars.
• Previously hidden features like AI summaries and full-content import are now visible as disabled options.

Add sources like RSS feeds, GitHub repositories, and YouTube channels directly from the feed menu. Import content from files or other apps without navigating to settings.

Sources can now automatically apply tags based on specific patterns like URLs or keywords. This helps organize content before AI tagging runs.

• Tag items from specific GitHub repositories or other sources using simple rules.
• Define tags based on URL, title, notes, or content.

Users can now set default tags for their sources to help AI agents organize and distinguish content from different feeds.

• Sources like X bookmarks and GitHub stars now come with pre-configured tags to improve organization.
• Default tags are applied automatically during ingestion while still allowing AI auto-tagging to run.

Users can now receive notifications when items change and sync data efficiently.

• Receive real-time notifications for item changes.
• Sync data incrementally to save bandwidth.