Skip to content

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

0.3.1-alpha.1 - 2026-04-01

Patch release fixing 13 config field mismatches found during QA testing.

Fixed

  • IdentitySettings: use format/aieos fields instead of fabricated name/description
  • ObservabilitySection: backend, otel_endpoint, runtime_trace_mode (5 wrong fields replaced)
  • RuntimeSection: kind, reasoning_effort (removed streaming/max_tokens/timeout)
  • SecretsSection: encrypt only (removed 5 fabricated vault fields)
  • TunnelSection: nested provider-specific configs (ngrok/cloudflare/tailscale)
  • WebSearchSection: brave_api_key, searxng_instance_url (removed generic api_key/safe_search)
  • HeartbeatSection: interval_minutes not interval_secs
  • Mission analysis: use zeroclaw CLI instead of direct API calls (fixes encrypted key 401)
  • JSON extraction: brace-matching fallback for CLI response parsing
  • CI workflow: --legacy-peer-deps and xvfb-run for E2E

0.3.0-alpha.1 - 2026-03-31

Third alpha release — upgraded to zeroclaw 0.6.7 with full QA pass, config corrections, and new features for memory, auth, gateway, and diagnostics.

Added

New Pages

  • Memory management page: list, get, stats, clear with category/session filtering and pagination
  • SOP page with graceful "not available" fallback for CLI versions without SOP support

New Settings Sections (28 total)

  • Authentication: manage provider auth profiles (list, status, logout)
  • Triggers: webhook endpoints, rate limiting, MQTT broker configuration
  • Query Routing: classification rules with hint/keywords/priority
  • Sub-Agents: delegation configuration with provider/model/tools per agent
  • Skills Config: open-skills toggle, directory, prompt injection mode
  • Composio: OAuth tool integration
  • HTTP Requests: domain allowlist, size limits, timeout
  • Peripherals: board configuration with transport/path/baud
  • Security: OTP, E-Stop, sandbox, resource limits (corrected field names)

zeroclaw 0.6.7 Integration

  • E-Stop: engage/status/resume with confirmation dialog and sidebar button
  • Diagnostics tab: doctor traces, model validation, self-test (quick/full)
  • Metrics dashboard widget
  • Service logs tailing with line count selector
  • Skills test runner and config schema viewer
  • Gateway management: restart, get/generate pairing codes
  • Agent --peripheral flag for hardware attachment in chat
  • CLI update check in General settings

Error Handling & UX

  • React Error Boundary wrapping entire app
  • Global error handlers in main and renderer processes
  • Focus management on route changes
  • Improved empty states for skills and missions pages
  • Onboarding: channel skip guidance, "What's next?" section

Testing & Quality

  • 1,180+ unit tests (up from 849)
  • Tests for all 28 settings sections
  • Tests for TagInput, ConfirmDialog, DiagnosticsTab, MemoryPage, AuthSection
  • Sidebar collapse persistence tests

Performance & Security

  • React.memo on BudgetWidget, MetricsWidget, DiagnosticsTab, EstopButton
  • CSP headers refined: img-src, font-src, connect-src
  • Context menu disabled in production
  • webSecurity explicitly set

Fixed

  • Settings crash: removed useBlocker (requires data router, incompatible with HashRouter)
  • ANSI escape codes: stripped globally from all CLI output in runCommand()
  • Config field mismatches: SecuritySection (sandbox, resources), CostSection (USD not cents), BrowserSection (native_headless), StorageSection (provider) — all corrected to match 0.6.7 schema
  • Channel start: fixed to no-args call (0.6.7 starts ALL channels)
  • E-Stop disabled state: shows "Disabled" instead of error when not enabled in config

Changed

  • Upgraded zeroclaw CLI requirement from 0.1.0 to 0.6.7
  • Channel types expanded to 19 (added Nextcloud Talk, Linq, Nostr, MQTT, Feishu)
  • Chat overrides panel: added peripheral selector

0.2.0-alpha.1 - 2026-03-15

Second alpha release with major new features, security hardening, developer tooling, and expanded platform integration.

Added

Missions

  • Mission CRUD with JSON persistence
  • LLM-powered mission analysis with contribution planning
  • Workspace contribution applier with marker tracking
  • Mission detail page with overview and configuration tabs

Settings (19 sections, +3 new)

  • Heartbeat section: interval and timeout configuration
  • Tunnel section: provider, auth token, subdomain
  • Service section: OS service install/start/stop/uninstall management
  • Config export and import with security warnings
  • Unsaved changes warning dialog on navigation
  • Input validation for host, port, model ID, and required fields
  • Confirmation dialogs for destructive actions

Dashboard

  • Budget tracking tab with daily/monthly cost visualization
  • Daemon status banner in chat with start button

Hardware

  • Hardware discovery page with USB device scanning
  • Peripheral listing and management

System Integration

  • zeroclaw:// deep link protocol with path traversal protection
  • OS-level system notifications for background events
  • Auto-update mechanism via electron-updater (GitHub Releases)
  • Keyboard shortcuts (Cmd+S save, Cmd+, settings)
  • Quick navigation items in system tray menu
  • Window size/position persistence across restarts

Security

  • Path traversal vulnerability fix in workspace file operations
  • Input validation for cron operations to prevent command injection
  • Restrictive file permissions on config and workspace files
  • MCP app_eval tool gated behind ZEROCLAW_MCP_ALLOW_EVAL env var

Accessibility

  • ARIA labels on sidebar navigation and collapse button
  • ARIA live regions for daemon status, save feedback, and chat banner
  • Icons on daemon status badges for color-blind users

Performance

  • Dashboard refresh cooldown to prevent redundant API calls
  • Memoized log filtering to avoid refiltering on every render
  • Marketplace cache freshness check before fetching catalog
  • Dashboard timestamp re-render reduced from 1s to 30s interval

Developer Experience

  • ESLint 9 with React, hooks, a11y, and TypeScript plugins
  • Prettier with husky pre-commit hooks and lint-staged
  • type-check, test:coverage, and validate npm scripts
  • CI pipeline: lint, type-check, format-check, and coverage checks
  • Automated accessibility testing with vitest-axe

Testing

  • 849+ unit tests with expanded coverage
  • Tests for AI assist IPC, tray module, missions, chat daemon status, hardware, and more

Fixed

  • Updated zeroclaw repo links from openagen to zeroclaw-labs
  • Synced latest zeroclaw documentation

0.1.0-alpha.1 - 2026-02-25

First alpha release of ZeroClaw UI — a desktop management interface for the zeroclaw AI agent platform.

Added

Core Application

  • Electron 40 + React 19 desktop app with Vite-based build system
  • Preload script with typed IPC bridge and context isolation
  • System tray integration with daemon status indicator
  • Dark mode support with system theme sync
  • Single-instance enforcement
  • MCP server for Claude Code integration (daemon management, CDP debug tools)

Onboarding

  • 4-step setup wizard: detect installation, provider setup, channel config, test connection
  • First-run detection with automatic daemon startup

Dashboard

  • Operations dashboard with status, doctor, channels, scheduler, and logs tabs
  • Real-time daemon status polling (start/stop/restart)
  • Log viewer with real-time streaming
  • Channel health diagnostics

Settings (16 sections)

  • General (provider, model, temperature, API key)
  • Identity, Gateway, Memory, Channels, Model Routes
  • Scheduler with visual cron job builder
  • Autonomy, Runtime, Multimodal, Cost & Limits
  • Observability, Browser, Web Search, Storage, Secrets
  • SaveBar with TOML serialization, validation via zeroclaw doctor, and daemon restart
  • Window focus config reload for concurrent editing support

Agent Editor (7 sections)

  • Workspace file editor for Identity, Soul, Agent Definition, User, Memory, Heartbeat, Tools
  • Markdown editor with edit/view toggle
  • AI-assisted content generation dialog with user prompts
  • Dirty tracking with save bar

Skills Marketplace

  • Installed skills tab with remove functionality
  • Marketplace tab with search and category filtering
  • Bulk install with progress tracking and retry logic

Chat

  • Message-based chat interface for testing agent responses
  • Markdown rendering for agent replies
  • ANSI stripping from CLI output

Documentation Site

  • VitePress-powered docs site with GitHub Pages deployment
  • Setup guide and contributor information

Testing

  • 637+ unit tests (Vitest) with 87% code coverage
  • 9 Playwright E2E test files covering all pages and key interactions
  • Coverage thresholds: 85% statements/lines, 80% branches, 60% functions

CI/CD

  • GitHub Actions pipeline for unit tests and E2E tests on PRs and pushes to main/develop
  • Automated docs site deployment to GitHub Pages

Known Issues

  • See open issues for the full list
  • Emergency stop (E-Stop) UI not yet implemented
  • SOP management not yet in UI
  • 6 channel types not yet configurable (Nextcloud Talk, Linq, Nostr, MQTT, Feishu, CLI)

Released under the Business Source License 1.1.