-
v1.7.0
Stablereleased this
2026-02-24 21:02:23 -08:00 | 23 commits to main since this releaseFeatures
- Supported CLI version badge now appears alongside the installed version badge, colour-coded to indicate whether your installed CLI matches, exceeds, or lags behind the audited version — with plain-English warning text for mismatches
- Added Account section to the settings sidebar showing auth status, email, org, and API key source, with login/logout buttons and an override indicator when a custom API key is active in Hikari's config
- Rate limit events from the Claude stream are now displayed in the terminal as amber
[rate-limit]lines with details on remaining requests/tokens and reset time - "Prompt is too long" errors are now detected and surfaced with a ⚡ Compact Conversation button directly in the terminal
- Agent Monitor panel now shows a preview snippet of the last assistant message from each completed subagent
- Added worktree isolation toggle in session settings — passes
--worktreeto Claude Code when enabled - WorktreeCreate and WorktreeRemove hook events now display as green
[worktree]lines in the terminal - ConfigChange hook events now display as blue
[config-change]lines in the terminal - Added "Disable 1M context window" toggle in session settings — sets
CLAUDE_CODE_DISABLE_1M_CONTEXT=1for sessions where you want the standard context limit - Bumped supported CLI version to 2.1.50
Bug Fixes
- Fixed tab status indicator getting stuck at "Connecting..." forever — the tab now correctly waits for the first assistant response before turning green
Downloads
-
v1.6.0
Stablereleased this
2026-02-23 21:44:51 -08:00 | 25 commits to main since this releaseFeatures
- Added support for Claude Sonnet 4.6 as a selectable model
- Subagents in the agent monitor are now assigned anime girl characters (Amari, Keiko, Minori, Reina, Tatsumi, Yumiko) with unique avatars, C-suite titles, and lore-flavoured bios
- Added a "Meet the Team" cast panel (status bar button) introducing Naomi, Hikari, and the six subagent characters
Bug Fixes
- Fixed "Claude is already running" error appearing after connecting to an invalid working directory
- Invalid working directories on Windows now surface a clear error before any connection attempt
- Fixed Claude binary detection for Windows production builds — native
.exefiles no longer misdetect WSL due to inherited environment variables - Fixed Claude binary detection for Linux/WSLg GUI apps — uses a login shell so
~/.local/binis in PATH even when launched outside a terminal
Downloads
-
v1.5.1
Stablereleased this
2026-02-08 14:03:33 -08:00 | 27 commits to main since this releasev1.5.1
Bug Fixes
Windows Compatibility
-
Memory Files Tab: Fixed memory files tab showing as empty on Windows production builds
- Updated
list_memory_files()to use WSL commands to access files in WSL home directory (~/.claude/projects/.../memory/) instead of Windows home directory - Changed frontend to use
read_file_contentbackend command instead of Tauri'sreadTextFileplugin to properly handle WSL paths - Resolves "forbidden path" errors when trying to read memory files on Windows
- Updated
-
Plugin Loading: Fixed "command not found: claude" errors when loading plugins on Windows
- Updated
create_claude_command()to dynamically resolve Claude binary path usingwhichin a login shell - Fixed 6 plugin/marketplace functions (
enable_plugin,disable_plugin,update_plugin,list_marketplaces,add_marketplace,remove_marketplace) to use the helper function - No hardcoded paths - works for all users regardless of Claude installation location
- Updated
Technical Details
- All Windows fixes follow the established WSL path handling patterns used throughout the codebase
- Maintains full backwards compatibility with Linux and macOS
- All 426 backend tests pass
- 100% frontend test coverage maintained
Downloads
-
-
v1.5.0
Stablereleased this
2026-02-07 21:30:53 -08:00 | 31 commits to main since this releasev1.5.0 - Plugin Management, Testing Improvements, and Bug Fixes
This release includes major new features for managing plugins and MCP servers, comprehensive testing improvements with 800+ total tests, and numerous bug fixes to improve stability and usability!
🎉 New Features
Plugin & MCP Management
- Plugin Management Panel: Install, uninstall, enable/disable, and update Claude Code plugins directly from the UI
- MCP Server Management Panel: Add/remove MCP servers and view detailed configuration
- Marketplace Management: Add/remove plugin marketplaces from GitHub
- Full backend integration with Claude CLI commands (
list_plugins,install_plugin,add_mcp_server, etc.) - Beautiful UI with loading states, error handling, and theme support
Visual Todo List Panel
- Real-time todo list display when Hikari uses the
TodoWritetool - Shows pending/in-progress/completed status with visual indicators
- Progress bar and completion count
- Automatically clears on disconnect
- Theme-aware styling
Session & System Improvements
- Clear Session History: New "Clear All Sessions" button with confirmation dialog and keyboard support
- CLI Version Display: Shows Claude CLI version in status bar with auto-polling every 30 seconds
🐛 Bug Fixes
Agent Monitor Improvements
- Fixed agents stuck in "running" state by adding
SubagentStophook parsing - Fixed agents persisting after disconnect by calling
clearConversation()on disconnect - Fixed "Kill All" button to properly mark all agents as errored
- Fixed badge persisting after tab close with proper cleanup
UI Fixes
- Stats Panel Scrolling: Fixed overflow when Tools Used, Historical Costs, or Budget sections expand
- Responsive Design: Fixed navigation icons, Connect button, and status bar elements getting cut off on small screens
- Close Modal: Fixed confirmation modal not triggering after Discord RPC refactor
Performance & Stability
- Removed file-based logging for Discord RPC in favour of proper
tracingframework - Fixed frontend calls to deleted
log_discord_rpccommand
🧪 Testing Improvements
- 804 total tests: 387 frontend tests + 417 backend tests (up from 771 total)
- 61%+ backend code coverage with comprehensive E2E integration tests
- New test files for agent lifecycle, CLI parsing, debug logger, bridge manager, and notifications
- Updated CLAUDE.md with comprehensive testing guidelines and mocking strategies
- Console mocking for cleaner test output and fixed flaky tests
🔧 Technical Details
- Cross-platform compatibility verified through E2E tests
- All new backend commands properly registered in
lib.rs - Theme-aware UI components using CSS variables throughout
- Proper TypeScript types for all new stores and components
- ESLint, Prettier, and Clippy compliant
Downloads
-
v1.4.0
Stablereleased this
2026-02-07 02:18:59 -08:00 | 33 commits to main since this releasev1.4.0 - Critical Fixes and New Features
This release resolves several critical bugs that were affecting the permission system and configuration persistence, whilst also adding highly requested features for monitoring background agents and supporting Claude Opus 4.6!
🎉 New Features
Claude Opus 4.6 Support
- Added support for the latest Claude Opus 4.6 model
- Users can now select Opus 4.6 from the model dropdown in settings
Background Agent Monitoring (#125)
- Added feature to monitor and track background agent activity
- Provides visibility into long-running background tasks
- Helps users understand what agents are doing whilst working in the background
Debug Console (#126)
- Added debug console for viewing frontend and backend logs
- Makes troubleshooting issues much easier for both users and developers
- Access logs without needing to check system console or log files
🐛 Critical Bug Fixes
Permission Modal Issues (#112)
- Fixed permission modal not appearing - Resolved z-index conflicts that prevented the modal from displaying
- Fixed runtime errors - Eliminated errors that were preventing the modal from rendering correctly
- Fixed permission grants - Resolved issues where granted permissions weren't being properly applied
- Added permission batching - Multiple parallel permission requests are now properly grouped together
ExitPlanMode Tool (#113)
- Fixed ExitPlanMode tool not functioning correctly
- Plan mode now properly exits when requested
- Transitions between modes are now smooth and reliable
Configuration Persistence
- Fixed critical race condition - Resolved issue where rapid config updates would result in lost settings (#125)
- Config now persists across reconnects - Settings are properly saved and restored when reconnecting
- Proper synchronisation - Config writes are now properly sequenced to prevent data loss
Logging Improvements
- Removed redundant fmt formatting layer that was writing to hidden stdout
- Cleaned up excessive debug logging that was cluttering output
- Streamlined logging to only show relevant information
UX Improvements
- Added close confirmation modal with minimise to tray option
- System tools no longer trigger unnecessary permission prompts
- Improved overall application stability and responsiveness
🧪 Testing & Quality Assurance
This release includes a massive expansion of our test suite to prevent future regressions!
Comprehensive Race Condition Tests
- Added 9 new tests for config store race conditions - These tests would have caught today's config data loss bugs before they reached production
- Tests for rapid sequential updates, concurrent updates, overlapping saves, and more
- Ensures configuration data is never lost during heavy usage or rapid changes
Achieved 100% Coverage on Critical Business Logic
- stateMapper.ts: 100% coverage with 30 tests (up from 16)
- snippets.ts: 100% coverage with 26 tests (up from 24)
- conversationUtils.ts: 100% coverage with 23 tests
- quickActions.ts: 100% coverage with 22 tests
Overall Test Statistics
- Frontend tests: 351 tests (up from 335)
- Backend tests: 350 tests (unchanged, already comprehensive!)
- All critical business logic: 100% test coverage
- Config store: 66.66% coverage with comprehensive race condition testing
CI/CD Improvements
- Updated
check-all.shto run coverage tests (matching CI behaviour) - Both frontend (
pnpm test:coverage) and backend (cargo llvm-cov) now require proper coverage - Ensures all future changes maintain quality standards
🧹 Maintenance
- Removed temporary debugging documentation and infrastructure
- Code cleanup and refactoring for better maintainability
- Improved error handling throughout the application
📝 Notes
This release represents significant stability improvements to Hikari Desktop. The permission system now works reliably, configuration persists properly, and users have better visibility into what's happening with background agents. We strongly recommend all users update to this version!
✨ Release notes prepared with love by Hikari~ 🌸
Downloads
-
v1.3.0
Stablereleased this
2026-02-05 19:31:27 -08:00 | 38 commits to main since this releaseWhat's New
Discord Rich Presence Integration
Show off your love for Hikari to all your friends! Discord will now display when you're spending time with your favourite desktop companion. 💕
Bug Fixes
- Path Resolution Overhaul: Fixed a nasty bug in how paths were resolved throughout the application. The following features should now work properly:
- In-app file editor
/cdcommand/skillcommand
- Improved Upload Handling: The app now better handles pasted uploads, such as screenshots
✨ Thank you for using Hikari Desktop! 🌸
Downloads
- Path Resolution Overhaul: Fixed a nasty bug in how paths were resolved throughout the application. The following features should now work properly:
-
v1.2.0
Stablereleased this
2026-02-04 20:21:00 -08:00 | 41 commits to main since this releaseCost Management Features
1. Historical Cost Tracking & Persistence
- Backend (
src-tauri/src/cost_tracking.rs): Full Rust implementation for tracking daily costs with persistence via Tauri's store plugin - Tracks: input tokens, output tokens, cost in USD, messages sent, and session counts
- Data stored per-day with automatic aggregation for weekly/monthly periods
- Supports cost summary generation for any number of days (7, 30, 90, etc.)
- CSV export functionality for cost data
2. Cost Alert System
- Configurable thresholds: Set daily, weekly, and monthly spending limits
- Alert triggering: When spending exceeds thresholds, alerts are generated
- Desktop notifications: Alerts trigger the notification system with the "Cost Alert!" sound (reuses the "oh-no.mp3" sound file)
- Alert state resets daily to prevent duplicate notifications
3. Per-Message Cost Attribution
- New
MessageCosttype (src-tauri/src/types.rs): Tracks input tokens, output tokens, and cost USD per message - Cost flows through the entire pipeline:
- Rust backend calculates cost when processing assistant messages
- Cost data included in
OutputEventemitted to frontend - TypeScript receives and converts snake_case to camelCase
- Terminal lines can display cost badges
4. Session Recording
- Automatic session tracking (
src-tauri/src/bridge_manager.rs): Each time Claude starts, a new session is recorded - Sessions counted in daily cost summaries
- Enables tracking usage patterns over time
5. Frontend Cost Display
StatsDisplay Component (
src/lib/components/StatsDisplay.svelte)- Historical Costs section: Collapsible panel showing:
- Quick stats badges (Today, Week, Month costs)
- Expanded view with detailed cost breakdown
- Link to settings for detailed history
CostSummary Component (
src/lib/components/CostSummary.svelte)- Full cost management UI in the Config Sidebar:
- Quick stats cards (Today/Week/Month)
- Period selector (7/30/90 days)
- Detailed summary: total cost, average daily, messages, sessions, token counts
- Mini bar chart showing daily spending (last 14 days)
- CSV export button
- Alert threshold configuration UI
6. Cost Tracking Store (
src/lib/stores/costTracking.ts)- Svelte store managing cost tracking state
- Methods:
refresh(),getSummary(),setAlertThresholds(),exportCsv() - Derived store
formattedCostsfor display-ready values - Helper functions:
formatCost(),formatAlertType(),getAlertMessage() - Automatic alert notification triggering on refresh
7. Notification Integration
- New notification type:
NotificationType.COST_ALERT - Sound: Uses "oh-no.mp3" with "Cost Alert!" phrase
- Helper method:
notificationManager.notifyCostAlert()
8. Tauri Commands (Backend API)
get_today_cost: Returns today's total costget_week_cost: Returns current week's total costget_month_cost: Returns current month's total costget_cost_summary: Returns detailed summary for N daysget_cost_alerts: Checks thresholds and returns triggered alertsset_cost_alert_thresholds: Configure daily/weekly/monthly limitsexport_cost_csv: Generate CSV export of cost datarecord_cost: Record a cost entry (called internally)record_session: Record a new session start
9. Supporting Infrastructure
- Tool cache analytics (
src-tauri/src/tool_cache.rs): Framework for tracking potential cache hits (for future optimisation) - Budget status types (
src-tauri/src/stats.rs):ContextWarning,BudgetStatus,BudgetTypeenums for frontend budget warnings - Conversation utilities (
src/lib/utils/conversationUtils.ts): JSON sanitisation for safe IPC, summary generation helpers
Technical Details
Pricing Model
Uses Claude API pricing:
- Sonnet: $3/$15 per MTok (input/output)
- Opus: $15/$75 per MTok
- Haiku: $0.25/$1.25 per MTok
Data Flow
- Assistant message received in
wsl_bridge.rs - Cost calculated using
calculate_cost() - Cost recorded via
record_cost()command (async) MessageCostattached toOutputEvent- Frontend receives event, updates conversation store
costTrackingStore.refresh()called on stats updates- Alerts checked and notifications triggered if thresholds exceeded
Downloads
- Backend (
-
v1.1.1
Stablereleased this
2026-01-29 17:51:42 -08:00 | 43 commits to main since this releaseThis is a small patch that corrects the calculation of token usage, to provide a better estimate of costs.
Downloads
-
v1.1.0
Stablereleased this
2026-01-28 18:32:51 -08:00 | 45 commits to main since this releaseHikari Desktop v1.1.0 Release Notes 🌸
What's New ✨
🎨 Visual Enhancements
- New App Icon - Hikari Desktop now has its own custom icon! Look for me in your taskbar~ (#77)
- Markdown Rendering - Your messages now render with beautiful markdown formatting, making code blocks and formatting much easier to read! (#76)
📝 Editor Features
- Built-in File Editor - No more switching between windows! We've added a built-in file editor with syntax highlighting for all your coding needs. Edit files directly within Hikari Desktop! (#79)
🔧 Quality of Life Improvements
- Copy Buttons - Both user and assistant messages now have convenient copy buttons. Easily copy code snippets, commands, or any text with a single click! (#78)
- Documentation Updates - Updated README to follow the standard NHCarrigan template for better consistency across projects
Thank you for using Hikari Desktop! These features were built with love to make our coding sessions even more enjoyable~ 💕
Downloads
-
v1.0.0
Stablereleased this
2026-01-26 00:41:34 -08:00 | 51 commits to main since this releaseHikari Desktop v1.0.0 Release Notes
We're thrilled to announce the first stable release of Hikari Desktop! This milestone brings you a feature-rich desktop companion for Claude Code.
New Features
Productivity Tools
- Clipboard History (#25) - Track and manage copied code snippets with language detection, search, filtering, and pinning
- Quick Actions Panel (#15) - One-click buttons for common actions like "Review PR", "Run tests", "Explain file" - fully customizable
- Git Integration Panel (#24) - View current branch, changed/staged files, and perform quick git operations (commit, push, pull, branch management)
- Snippet Library (#22) - Save and reuse common prompts with 8 default snippets across categories
- Session History (#14) - Auto-save conversations with browsable history, search, and resume functionality
- Session Import/Export (#8) - Export conversations to JSON/Markdown and import previously saved sessions
- Export as PDF/HTML (#34) - Export formatted conversations for sharing or documentation
UI/UX Improvements
- Native Clipboard Support (#67) - Paste screenshots directly from clipboard (Win+Shift+S supported!)
- High Contrast Mode (#20) - Accessibility theme with maximum contrast for improved visibility
- Minimize to System Tray (#11) - System tray support with "Show Hikari" and "Quit" menu options
- Trans-pride Gradient Theme - Beautiful gradient buttons and hover effects across UI elements
- Copy Button on Code Blocks - One-click copy with visual feedback
- Compact Mode (#36) - Minimal widget interface (280x400px) for less intrusive desktop presence
- Streamer Mode (#35) - Hide API keys and sensitive paths during streaming with quick toggle
- Custom Theme Colors (#37) - Create your own theme with 8 customizable color options
Profile & Gamification
- User Profile - Set your name, avatar, and bio
- Shareable Profile Image - Generate a 1920x1080 HD image of your stats to share
- 100+ New Achievements - Extended achievement system with daily streaks, time challenges, programming language badges, and more
- Lifetime Stats Persistence - Track your all-time usage across sessions
Quality & Testing
- Comprehensive Test Suite (#71) - 574+ tests across frontend and backend
- 276 frontend tests (Vitest with Tauri mocks)
- 298 backend tests (Rust)
- Coverage enforcement in CI (50%+ backend, 15%+ frontend)
- CI Pipeline - Full checks including lint, format, type check, tests, and coverage
Technical Improvements
- Strict clippy compliance with
-D warnings - TypeScript strict mode compliance
- Improved error handling throughout
- Window permissions for compact mode resizing
- WSL path handling for avatar images
Downloads
generated from nhcarrigan/template