v0.4.0

What's New

Agent Controller

Multi-agent orchestration system inspired by claude-code-controller, adapted to OpenClawd's multi-provider architecture.

• Agent Management — Spawn, list, kill, and communicate with named agents via REST API
• Task Tracking — Create tasks with dependency tracking (blockedBy/blocks), assign to agents
• Session Management — Group agents into sessions with shared defaults (provider, model)
• Inter-Agent Messaging — Direct messages and broadcast between agents
• SSE Streaming — Stream agent responses in real-time via Server-Sent Events
• Desktop UI — New "Agents" tab in settings with spawn form, agent cards, task board, and session controls
• Permission Presets — full, edit, plan, ask — fail-closed on unknown presets
• 18 REST API Endpoints under /api/agents

Security

• XSS prevention with escapeAttr for HTML attributes and CSS.escape for selectors
• Reserved agent name validation (ask, tasks, sessions)
• Task status validation against enum
• Ephemeral agent name collision prevention

Tests

• 143 tests passing (71 new agent tests)

Full Changelog: v0.3.0...v0.4.0

What's Changed

• feat: add agent controller for multi-agent orchestration by @rohitg00 in #14

Full Changelog: v0.3.0...v0.4.0

v0.3.0

What's New

Token Usage Dashboard

• Real-time token count and cost tracking per message
• 7-day usage history with CSS bar chart
• Per-provider breakdown table (tokens in/out, cost, requests)
• New Settings > Usage tab

Model Fallback Router

• Automatic failover when primary provider errors (rate limit, auth, network)
• Configurable fallback provider and model in Settings > General
• SSE toast notification on fallback switch
• Supports all 20+ providers

Channel Bot Integration

• Telegram - polling bot with Markdown formatting, group mention support
• Discord - DM and @mention responses, 2000-char chunking
• WhatsApp - QR code pairing via Baileys, session persistence
• Slack - Socket Mode bot with DM and app_mention support
• Per-user conversation sessions with /reset, /model, /status commands
• New Settings > Channels tab with start/stop toggles per platform

Voice Input (Speech-to-Text)

• Browser Web Speech API integration (no external dependencies)
• Mic button in home and chat input areas
• Real-time interim transcript overlay
• Keyboard shortcut: Cmd+Shift+V (Mac) / Ctrl+Shift+V (Windows/Linux)

3D Animated Logo

• Claw mascot logo in desktop app sidebar with 3D floating animation
• Animated logo on landing page nav bar and hero section
• Hover bounce effect with 360-degree flip

Code Quality

• 72 tests (security, providers, MCP loader)
• 4 rounds of CodeRabbit review fixes (~49 issues resolved)
• XSS prevention: escapeHtml/escapeAttr, DOMPurify with SRI
• Command injection prevention in MCP loader
• CI workflow with GitHub Actions

Dependencies Added (server)

• node-telegram-bot-api - Telegram bot
• discord.js - Discord bot
• @whiskeysockets/baileys - WhatsApp Web
• @slack/bolt - Slack bot

Full Changelog: v0.2.0...v0.3.0

What's Changed

• fix: download buttons now resolve to direct binary URLs by @rohitg00 in #6
• docs: add detailed architecture diagram to README by @rohitg00 in #7
• Revert "docs: add detailed architecture diagram to README" by @rohitg00 in #8
• docs: add detailed architecture section to README by @rohitg00 in #9
• feat: add token dashboard, model fallback, channel bots, and voice input by @rohitg00 in #11
• feat: add 3D animated logo to sidebar and home view by @rohitg00 in #12
• feat: add 3D animated logo to desktop app and landing page by @rohitg00 in #13

Full Changelog: v0.2.0...v0.3.0

v0.2.0

What's New

CLI Onboarding Wizard (PR #4)

• First-run interactive setup via @clack/prompts
• Configure API keys (Anthropic, OpenAI, Gemini) on first launch
• Select MCP servers to enable (filesystem, memory, fetch, brave-search)
• npx openclawd-cli starts the server headlessly

Desktop Onboarding (PR #4)

• Full-screen onboarding overlay on first launch
• 3-step wizard: Welcome → API Keys → MCP Setup
• Skip option on each step, persists via localStorage

OS-Specific Downloads (PR #3)

• Website download buttons resolve to architecture-specific release assets
• Apple Silicon recommended by default on Mac, Intel shown as alternative
• Linux buttons include arch qualifiers (x86_64 AppImage, amd64 deb)
• GitHub API response cached in sessionStorage (10min TTL)
• response.ok check before parsing API response

npm Package (PR #5)

• Published as openclawd-cli on npm (40KB tarball)
• Install: npx openclawd-cli
• Binary name is openclawd

Install

npx openclawd-cli

Or clone:

git clone https://github.com/rohitg00/openclawd
cd openclawd && ./setup.sh

What's Changed

• Add architecture diagram to README by @rohitg00 in #2
• feat: add onboarding wizard and npx CLI entry point by @rohitg00 in #4
• Add OS-specific download links to website by @rohitg00 in #3
• feat: v0.2.0 release - openclawd-cli npm package by @rohitg00 in #5

Full Changelog: v0.1.0...v0.2.0

v0.1.0 - Alpha Release

OpenClawd v0.1.0 - Alpha Release

The first public release of OpenClawd — an open source AI desktop agent that unifies 20+ LLM providers, 20+ MCP integrations, and multi-platform messaging into a single app.

Highlights

• Cross-platform desktop app (macOS, Windows, Linux)
• 20+ LLM providers with auto-discovery
• 20+ MCP server integrations from built-in catalog
• Messaging bot support (WhatsApp, Telegram, Signal, iMessage)

LLM Providers

Provider	Notable Models
Anthropic Claude	Opus 4.5, Sonnet 4, 3.5 Sonnet, 3.5 Haiku
Claude Pro/Max	Subscription-based access
OpenAI	GPT-4o, o1, o1-mini, GPT-4 Turbo
Google Gemini	2.0 Flash, 1.5 Pro (up to 2M context)
Groq	Llama 3.3 70B (ultra-fast inference)
Mistral	Large, Codestral, Mixtral 8x22B
DeepSeek	Chat, Coder, R1 (reasoning)
xAI Grok	Grok 2, Grok 2 Mini
OpenRouter	200+ models via unified API
Ollama	Auto-discovered local models
Amazon Bedrock	Claude, Titan, Llama via AWS
Together AI	Llama 3.1 405B, Mixtral
Fireworks	High-throughput inference
Perplexity	Search-enabled Sonar models
Cerebras	Llama 3.1 70B/8B
Venice AI	Privacy-focused models
Moonshot	128K context
MiniMax	245K context
Cloudflare AI	Free Llama models
GitHub Copilot	Via GitHub token

MCP Integrations (Built-in Catalog)

Install from the catalog with one click — no config files needed.

No Auth Required: Filesystem, Memory, Fetch, Git, Time, Sequential Thinking

Developer Tools: GitHub, GitLab, Linear

Communication: Slack, Discord

Productivity: Google Drive, Notion

Search: Brave Search, Exa

Databases: PostgreSQL, SQLite

Cloud: AWS (S3, Lambda, EC2)

Browser Automation: Puppeteer, Playwright

Desktop App Features

• Multi-chat management with persistent history
• Per-chat provider and model selection
• Streaming responses with abort capability
• File attachments (images + text, up to 5 files)
• Markdown rendering with syntax highlighting
• Thinking mode toggle (normal/extended)
• Inline tool call visualization with input/output
• Task tracker sidebar (from TodoWrite)
• Live browser embed panel
• Settings UI for API keys and MCP servers
• MCP catalog browser with category filtering

Messaging Bot

Run OpenClawd as a bot on messaging platforms:

• WhatsApp — QR code auth, group + DM support, image handling
• Telegram — Bot token auth, markdown formatting, message chunking
• Signal — Phone number-based auth via signal-cli
• iMessage — Native Apple protocol (macOS only)

Downloads

Platform	File
macOS (Apple Silicon)	OpenClawd-1.0.0-mac-arm64.dmg
macOS (Intel)	OpenClawd-1.0.0-mac-x64.dmg
Windows (x64)	OpenClawd-1.0.0-win-x64.exe
Linux (x64)	OpenClawd-1.0.0.AppImage
Linux (Debian)	openclawd_1.0.0_amd64.deb

Note: Binaries are unsigned in this alpha release. On macOS, right-click and select "Open" to bypass Gatekeeper. On Windows, click "More info" then "Run anyway" in SmartScreen.

Quick Start (from source)

git clone https://github.com/rohitg00/openclawd
cd openclawd
npm install
npm run dev

What's Changed

• feat: cross-platform binary builds and Vercel deployment by @rohitg00 in #1

Full Changelog: https://github.com/rohitg00/openclawd/commits/v0.1.0