Integrations

Connect Ultrathink with your favourite tools and services.

Integrations

Ultrathink connects with your browser, desktop, and AI services to capture and enhance your knowledge.

Browser extension

The extension is the fastest way to capture content from any webpage. It runs as a Manifest V3 Chrome extension with support for Edge and other Chromium browsers.

Installation

  1. Visit the Chrome Web Store
  2. Click Add to Chrome
  3. Pin the extension to your toolbar for quick access

Note: The latest version can always be downloaded from the web app at tryultrathink.com/downloads/ultrathink-extension.zip. To install manually:

  1. Download and extract the zip file
  2. Open chrome://extensions
  3. Enable Developer mode
  4. Click Load unpacked and select the extension folder

Setup

The extension uses a device key to link captures to your account.

  1. Sign in to tryultrathink.com
  2. Go to Settings and find Device Keys
  3. Click Generate new key
  4. Copy the key (it is only shown once)
  5. Right-click the extension icon and select Options
  6. Paste your device key and save

A green indicator appears when the key is valid.

Capturing content

Click the extension icon or press Ctrl+Shift+4 (Mac: Cmd+Shift+4) to open the capture popup.

What gets captured:

  • Page URL and title
  • Selected text (if any)
  • Tab group name and colour (if the tab is in a group)
  • Page metadata (description, Open Graph image, author, estimated reading time)
  • Your notes (optional)

Auto-save: The popup starts a 3-second countdown. Any interaction (typing, changing options) resets the timer. When the timer ends or the popup closes, the entry is saved automatically.

Bulk capture: Check "All tabs" to save every open tab in the current window. Each tab is saved as a separate entry with a "saved-together" relationship linking them.

Screenshots: Press Ctrl+Shift+5 (Mac: Cmd+Shift+5) to capture a screenshot.

  • Click and drag to select a region
  • Single click to capture the full visible page
  • Press Escape to cancel

Screenshots are compressed to JPEG (70% quality, max 1920px width) to stay within Firestore's 1MB document limit.

The extension automatically identifies content types from URLs.

PatternTypeExamples
chat.openai.com, chatgpt.comChatGPTAI conversation threads
claude.aiClaudeClaude conversation threads
perplexity.aiPerplexityPerplexity AI searches
outlook.office.com, outlook.live.comOutlookEmails (converted to deep links)
sharepoint.com with :w:MS WordSharePoint/OneDrive documents
sharepoint.com with :x:MS ExcelSharePoint/OneDrive spreadsheets
sharepoint.com with :p:MS PowerPointSharePoint/OneDrive presentations
sharepoint.com with :b: or onenote.aspxMS OneNoteOneNote notebooks
notion.so, notion.siteNotionNotion pages and databases
youtube.com, vimeo.comVideoVideo content
URLs ending in .pdfPDFPDF documents
URLs ending in .mdMarkdownMarkdown files

Outlook handling: Outlook web URLs are transformed into deep links that open the specific email rather than redirecting to the inbox.

AI conversation extraction: For ChatGPT and Claude, the extension extracts the full conversation thread (user and assistant messages) and stores it in the entry for AI processing.

Keyboard shortcuts

ActionWindows/LinuxMac
Open capture popupCtrl+Shift+4Cmd+Shift+4
Screenshot captureCtrl+Shift+5Cmd+Shift+5

Customise shortcuts at chrome://extensions/shortcuts.

Architecture and security

Manifest V3: The extension uses the latest Chrome extension architecture with a service worker instead of a persistent background page.

Permissions:

  • activeTab: Access the current tab for capture
  • tabs: Query open tabs for bulk capture
  • tabGroups: Read tab group metadata
  • scripting: Inject content scripts for text extraction
  • storage: Store device key and settings

Security measures:

  • Device key is SHA-256 hashed before transmission
  • Full key never leaves your device
  • Content scripts are isolated from page JavaScript
  • No data is logged or transmitted beyond Firebase

Desktop widget

The Electron desktop widget keeps capture always available, even when your browser is closed.

Installation

Windows:

  1. Download the installer from tryultrathink.com/downloads/ultrathink-setup.exe
  2. Run the .exe installer
  3. The widget starts automatically and appears in your system tray

macOS:

  1. Download the .zip file for your Mac:
  2. Extract and move Ultrathink.app to Applications
  3. Open the app (you may need to right-click and select Open the first time)

Setup

  1. Right-click the system tray icon
  2. Select Paste Device Key
  3. The key is read from your clipboard (generate one from Settings if needed)
  4. A notification confirms when the key is saved

Features

Drop zone: Drag and drop files directly onto the widget. Supported formats include PDF, Word, Excel, PowerPoint, images, audio, and video.

Notes: Type quick notes with rich text formatting (bold, italic, lists, links). Auto-save triggers after 3 seconds of inactivity.

Expanded mode: Click the expand button for a larger editor (380x480px) with a 10-second background save timer.

Screenshots:

  • Click the camera icon for full-screen capture
  • Double-click for region selection
  • Multi-monitor support: captures from the correct display

Audio recording:

  • Microphone: Click the mic icon to record voice notes
  • System audio (Windows only): Captures desktop audio for recordings

Always-on-top: The widget floats above other windows. Toggle via the system tray menu.

Start on login: Enable via the system tray menu to launch automatically.

Architecture and security

Electron framework: The widget runs as a desktop application with separate main (Node.js) and renderer (Chromium) processes.

Context isolation: The renderer process cannot access Node.js APIs directly. All system operations go through a secure IPC bridge with predefined channels.

Device key storage: Stored in the OS-level app data directory, not synced to cloud drives like OneDrive or iCloud.

Data flow:

  1. Content is captured in the renderer process
  2. Device key is hashed with SHA-256
  3. Entry is sent to the Firebase Cloud Function
  4. Cloud Function validates the hash and saves to Firestore

Permissions:

  • Screen capture (for screenshots)
  • Microphone (for audio recording)
  • Desktop audio (Windows only)
  • File system (for dropped files)

iOS (beta)

Ultrathink has a dedicated mobile interface when you browse the web app on iPhone or iPad. This is currently in beta.

How to use

  1. Open tryultrathink.com in Safari on your iPhone or iPad
  2. Sign in to your account
  3. The interface automatically adapts for mobile with a simplified capture menu

Features

The mobile interface provides quick capture options:

Notes: Create text notes with the keyboard Media: Capture photos, images, or files from your device Audio: Record voice notes with your microphone Paste: Save content from your clipboard

Add to Home Screen

For a more app-like experience, add Ultrathink to your home screen:

  1. Open tryultrathink.com in Safari
  2. Tap the Share button
  3. Select Add to Home Screen
  4. Tap Add

The app will open in full-screen mode without Safari's browser chrome.

Note: This is a progressive web app (PWA), not a native iOS app. Some features like background sync and push notifications may be limited compared to desktop

OpenAI integration

AI features in Ultrathink are powered by OpenAI models. This includes summarisation, classification, relationship detection, and natural language search.

How it works

When you save an entry, it goes through a multi-step AI pipeline:

  1. Title generation: If the title is generic ("Note", "Untitled", etc.), AI generates a descriptive title
  2. Summarisation: Content is summarised based on type (image, audio, document, link, or text)
  3. Classification: AI extracts entity type, topics, people, and work/personal category
  4. Relationships: AI finds mentions of other entries in your knowledge base
  5. Similarity: AI identifies semantically similar entries

Each step updates the entry with structured metadata that improves search and organisation.

Models used:

  • Entry processing: gpt-5-nano (fast, cost-effective)
  • AI Canvas: gpt-5.2 (advanced reasoning, multi-modal)
  • Transcription: Whisper (gpt-4o-transcribe)

Bring your own key

Free plan users provide their own OpenAI API key. Pro plan users get AI features included.

To add your key:

  1. Get an API key from platform.openai.com/api-keys
  2. Go to Settings > Integrations
  3. Paste your key (must start with sk-)
  4. Click Test key to validate
  5. Save

Your key is stored encrypted in Firestore and only used server-side in Cloud Functions. It is never exposed to the browser.

Upgrading to Pro: Pro plan (£9.99/month) includes AI features without needing your own key, plus a 30-day free trial.

Customising prompts

All AI prompts can be customised per user.

Go to Settings and find the prompt tabs:

  • AI Canvas: System prompt for the AI Canvas compose feature
  • AI Search: Prompt for natural language search

Changes are saved automatically and take effect on the next AI operation.

Prompts are resolved in order: your custom prompt > master prompt. If you delete your custom prompt, the system default is restored.

AI search lets you query your knowledge base in natural language.

How to use:

  1. Click the search icon or press Cmd+K / Ctrl+K
  2. Type a question like "What did I save about project budgets?"
  3. AI searches your entries and returns an answer with links to relevant items

Behind the scenes:

  • Your most recent entries are formatted and sent to the AI model
  • The AI returns an answer plus timestamps of matching entries
  • Results include both your private entries and org-shared entries (if applicable)

Configuration:

  • Default limit: 500 entries searched
  • Customise the limit in Settings > AI Search
  • Org-shared entries use 40% of the limit

Planned integrations

We are building integrations to pull context from external services during AI processing.

GitHub

Search your repositories for related issues, pull requests, and discussions. When you save an entry mentioning a project, AI can reference recent GitHub activity for richer summaries.

Notion

Search your Notion workspace for pages with similar topics. AI can cross-reference your Ultrathink entries with Notion documentation.

Fastmail

Search your email for conversations about the same subject. When you save a meeting note, AI can find related email threads for context.

Asana

Search your Asana tasks and projects for related work items. AI can link entries to active tasks automatically.

All planned integrations are:

  • Optional: Enable only the services you use
  • Secure: OAuth or API token authentication, stored encrypted
  • Read-only: We only search, never modify your external data

Watch the changelog for updates as these integrations become available.