ClawdAgent is the most comprehensive autonomous AI agent system ever built as open-source. It runs 24/7 across multiple platforms, with 14-layer security, 9 intelligence subsystems, 20 specialized agents, 35 tools, and 67 skills — all connected through a unified brain. It can autonomously manage Facebook, LinkedIn, Twitter, and TikTok accounts, browse the web visually with VNC streaming, make phone calls via Twilio + OpenAI Realtime, and publish to 9 social platforms on a daily schedule with AI-generated content.
<div align="center">
</div>
It doesn't just respond to commands — it thinks proactively, learns from interactions, evolves its own capabilities, secures itself against attacks, and manages complex multi-step workflows autonomously.
Think of it as your personal AI operations center: it manages your servers, writes your code, creates content, trades crypto, automates your phone, monitors your systems, publishes to 9 social platforms — all while learning and getting smarter over time.
You: "Check my servers, fix anything broken, and send me a summary on Telegram"
ClawdAgent: ✅ SSH'd into 3 servers → found nginx down on VPS2 → restarted it
→ checked SSL certs (2 expiring soon) → renewed them
→ generated AI image for the report header
→ sent you a full report on Telegram with recommendations
→ scheduled daily health checks via cron
→ learned the fix pattern for next time
You: "Create a marketing video about our product and publish it everywhere"
ClawdAgent: ✅ Generated script with AI → created video via Kie.ai
→ generated thumbnail via fal.ai (FLUX)
→ published to Twitter, Instagram, TikTok, LinkedIn, YouTube, Facebook
→ scheduled follow-up posts for peak engagement times
→ tracked publishing costs ($0.12 total)
You: "Set up DCA trading on Bitcoin — $50 daily, stop if it drops 20%"
ClawdAgent: ✅ Connected to Binance → configured DCA strategy → set risk limits
→ placed first buy order ($50 USDT → 0.00052 BTC)
→ scheduled daily recurring buys via cron
→ set 20% drawdown kill switch → activated portfolio tracker
→ will send Telegram alerts on every trade + daily P&L summary
You: "Research the top 5 competitors in our space and email me a report"
ClawdAgent: ✅ Searched Brave + scraped competitor websites with Playwright
→ analyzed pricing, features, tech stacks of 5 companies
→ generated a detailed comparison report (3 pages)
→ created competitive positioning matrix
→ emailed the report via Gmail with PDF attachment
→ saved findings to knowledge base for future reference
You: "My phone isn't syncing emails, fix it"
ClawdAgent: ✅ Connected to your Android via ADB → checked Gmail app status
→ found sync disabled after last OS update → re-enabled sync
→ cleared app cache (was 2.1 GB) → force-synced inbox
→ verified 47 new emails loaded → sent you confirmation on Telegram
→ created a skill "fix-email-sync" for next time
What Makes ClawdAgent Different
Most AI agents are reactive — you ask, they answer. ClawdAgent is autonomous:
It thinks on its own — The Proactive Thinker spots problems before you do. Server going down? SSL cert expiring? Budget running low? ClawdAgent alerts you before it becomes a crisis.
It creates its own tools — Need a new capability? ClawdAgent builds it at runtime. No waiting for the next release.
It spawns its own agents — The Meta Agent creates new specialized agents on demand. Need a "Legal Document Analyzer"? Ask once and it exists forever.
It evolves itself — The Evolution Engine runs improvement cycles: learns from failures, optimizes prompts, merges redundant agents, discovers new patterns.
It secures itself — 14 layers of Zero Trust defense. Every memory entry is SHA-256 checksummed. Every tool is hash-pinned. Every command is sandboxed. Every security check fails closed.
It runs your business — Not just chat. Crypto trading with 5 strategies. Social publishing to 9 platforms. Server management with auto-repair. Email, calendar, browser automation, mobile control.
It speaks every protocol — MCP (Anthropic), A2A (Google), ACP (IBM). Your agent can talk to any other compliant agent on the internet.
It runs anywhere — Linux VPS, Mac, Windows, Docker, Raspberry Pi. Local models (Ollama) for full privacy. No cloud dependency.
| Capability | AutoGPT | CrewAI | LangChain | MetaGPT | OpenDevin | ClawdAgent |
|-----------|---------|--------|-----------|---------|-----------|----------------|
| Autonomous Execution | Partial | Task-based | Chain-based | Workflow | IDE-focused | Full 24/7 autonomy |
| Persistent Memory | Limited | No | Via plugins | Partial | Session | PostgreSQL + Redis + Memory Hierarchy |
| Self-Evolution | No | No | No | No | No | Evolution Engine + Self-Repair + Auto-Learn |
| Proactive Thinking | No | No | No | No | No | Autonomous: spots problems, sends alerts |
| Multi-Platform Chat | No | No | No | No | No | Telegram + Discord + WhatsApp + Web + Voice/Phone |
| Web Dashboard | Basic | No | LangSmith | No | Browser | Full React Dashboard (26 pages) |
| Dynamic Tool Creation | No | No | No | No | No | Creates new tools at runtime |
| Multi-Agent Teams | No | Yes | No | Yes | No | Crew Orchestrator + Meta Agent + Factory |
| Agent-to-Agent Protocols | No | No | No | No | No | A2A + ACP + MCP (full compliance) |
Security Comparison
| Security Feature | AutoGPT | CrewAI | LangChain | MetaGPT | OpenDevin | ClawdAgent |
|-----------------|---------|--------|-----------|---------|-----------|----------------|
| Defense Layers | 1 | 1 | 2 | 1 | 2 | 14-layer defense-in-depth |
| Prompt Injection Guard | No | No | Basic | No | No | 20+ regex patterns + AI detection |
| Social Engineering Detection | No | No | No | No | No | 15 patterns, auto-block |
| Memory Integrity | No | No | No | No | No | SHA-256 checksums + tamper quarantine |
| Audit Trail | No | No | No | No | No | Tamper-evident hash chain |
| Command Sandboxing | No | No | No | No | Docker | Bash sandbox + command guard |
| RBAC + JWT Auth | No | No | No | No | No | Full RBAC + JWT + per-user perms |
| Kill Switch | No | No | No | No | No | Emergency stop + cost tracking |
| Approval Gate | No | No | No | No | No | Human-in-the-loop for critical ops |
| Governance Engine | No | No | No | No | No | Risk budgets + autonomy levels |
| Fail-Closed Design | No | No | No | No | No | All security checks fail-closed |
| Skill Scanning | No | No | No | No | No | 25+ patterns, severity scoring |
| Tool Integrity Hashing | No | No | No | No | No | SHA-256 hash-pinning on tools |
| Encryption at Rest | No | No | No | No | No | AES encryption + key rotation |
Intelligence & AGI Readiness
| Intelligence Feature | AutoGPT | CrewAI | LangChain | MetaGPT | OpenDevin | ClawdAgent |
|---------------------|---------|--------|-----------|---------|-----------|----------------|
| Multi-Model Support | GPT only | Multiple | Multiple | GPT-focused | Multiple | Claude 4.6 + OpenRouter (400+) + Ollama + Claude Code |
| Smart Model Routing | No | No | No | No | No | Auto-routes by complexity, cost, budget |
| Extended Thinking | No | No | No | No | No | 32K thinking tokens for deep reasoning |
| Intent Classification | Basic | No | No | Role-based | No | 45+ intents, multilingual (Hebrew, Arabic, CJK) |
| Intelligence Bridge | No | No | No | No | No | 9 subsystems: scoring, memory, governance, cost, routing, observability, goals, safety, feedback |
| Safety Simulator | No | No | No | No | No | Dry-run testing + impact assessment + rollback |
| Cost Intelligence | No | No | Callbacks | No | No | ROI tracking + budget forecasting + burn prediction |
| Behavior Adaptation | No | No | No | No | No | Multi-language personality + adaptive styles |
| Goal Engine | No | No | No | No | No | Self-initiated 30/60/90-day goals with KPIs |
| Feedback Loop | No | No | No | No | No | Pattern recognition + prompt optimization |
Tools & Integrations
| Category | AutoGPT | CrewAI | LangChain | MetaGPT | OpenDevin | ClawdAgent |
|----------|---------|--------|-----------|---------|-----------|----------------|
| Built-in Tools | 5-10 | 5-8 | Via plugins | 5-10 | 10-15 | 35 integrated tools |
| Pre-loaded Skills | No | No | No | No | No | 67 skills across 23 categories, extensible at runtime |
| Server Management | No | No | No | No | No | Multi-server SSH + Docker + health monitoring |
| Browser Automation | Via plugins | No | Via plugins | No | Built-in | Playwright headless + AI vision |
| Content Creation | No | No | No | No | No | AI video/image/music (Kie.ai 60+ models + fal.ai) |
| Social Publishing | No | No | No | No | No | 9 platforms (Twitter, IG, TikTok, LinkedIn, YT, FB, Threads, BlueSky, Pinterest) |
| Crypto Trading | No | No | No | No | No | 5 strategies + TA engine + risk manager |
| Mobile Automation | No | No | No | No | No | Android via ADB/Appium |
| Desktop Control | No | No | No | No | No | AI vision + mouse/keyboard |
| Email Integration | No | No | No | No | No | Gmail API + SMTP |
| Voice (TTS/STT) | No | No | No | No | No | ElevenLabs + Twilio + OpenAI Realtime (phone calls) |
| Remote Agent Bridge | No | No | No | No | No | OpenClaw (Ed25519 device auth) |
Deployment & Operations
| Feature | AutoGPT | CrewAI | LangChain | MetaGPT | OpenDevin | ClawdAgent |
|---------|---------|--------|-----------|---------|-----------|----------------|
| Production-Ready | Partial | Library | Library | Prototype | Beta | PM2 + Docker + nginx production stack |
| One-Command Install | No | pip | pip | pip | Docker | bash install.sh (interactive wizard) |
| Observability | Basic logs | No | LangSmith | No | No | Timeline events + tool heatmaps + error clustering |
| Cron Scheduling | No | No | No | No | No | Built-in cron + recurring tasks |
| Queue System | No | No | No | No | No | BullMQ (4 job types, 5 parallel, 20 queue) |
| Self-Repair | No | No | No | No | No | 9 known fix patterns + AI diagnosis |
| Config Management | .env | Code | Code | Code | Docker | Zod-validated .env + YAML hot-reload + Web UI |
| Open Source | Yes | Yes | Yes | Yes | Yes | Apache 2.0 (patent + trademark protection) |
ClawdAgent vs OpenClaw — The Definitive Comparison
OpenClaw (by Peter Steinberger) is the most well-known personal AI assistant. Here's how ClawdAgent compares on every dimension:
Architecture & Intelligence
| Feature | OpenClaw | ClawdAgent |
|---------|----------|----------------|
| Architecture | Single agent + skills | 20 specialized agents + Meta Agent + Crew Orchestrator |
| Agent Brain | Single LLM call | Intelligence Bridge: 9 subsystems (scoring, memory, governance, cost, routing, observability, goals, safety, feedback) |
| Multi-Agent Orchestration | No (single thread) | Crew Orchestrator + Agent Factory + dynamic spawn |
| Agent Self-Creation | No | Meta Agent creates new agents at runtime based on need |
| Dynamic Tool Creation | No | Creates new tools at runtime + auto-tool discovery |
| Self-Evolution | Skill creation via chat | Evolution Engine: autonomous improvement cycles + self-repair (9 fix patterns) + auto-learn |
| Proactive Thinking | Heartbeats | Proactive Thinker: spots problems, finds opportunities, sends alerts + 9 alert types |
| Intent Classification | Basic routing | 45+ intents with multilingual support (Hebrew, Arabic, CJK) |
| Extended Thinking | No | 32K thinking tokens for deep reasoning |
| Goal Engine | No | Self-initiated 30/60/90-day goals with KPIs and milestones |
| Behavior Adaptation | Persona onboarding | Multi-language personality variants + adaptive interaction styles |
| Smart Model Routing | Provider fallback | Cost-aware routing: complexity → model, cost → provider, budget → tier |
Security & Trust
| Security Feature | OpenClaw | ClawdAgent |
|-----------------|----------|----------------|
| Defense Architecture | Sandboxed execution | 14-layer defense-in-depth (Zero Trust) |
| Prompt Injection Guard | Basic filtering | 20+ regex patterns + AI detection + pre-AI message guard |
| Social Engineering Detection | No | 15 patterns, auto-block on high severity |
| Memory Integrity | No | SHA-256 checksums on every entry, tamper quarantine |
| Audit Trail | Logs | Tamper-evident hash chain on ALL operations |
| Governance Engine | No | Risk budgets, autonomy levels, cost/risk limits |
| Fail-Closed Design | No (permissive) | ALL security checks fail-closed on error |
| Kill Switch | No | Emergency stop + cost tracking + failure data |
| Approval Gate | No | Human-in-the-loop for critical ops (trading, social, destructive) |
| Skill Scanning | Basic trust | 25+ static analysis patterns, severity scoring, reputation tracking |
| Tool Integrity | No | SHA-256 hash-pinning detects MCP rug-pulls |
| RBAC + JWT Auth | Basic auth | Full RBAC + JWT + per-user permissions + rate limiting |
| Command Sandboxing | Docker sandbox | Bash sandbox + command guard + bounded concurrency |
| Encryption at Rest | No | AES encryption + automatic key rotation |
Bottom line: OpenClaw is an excellent personal AI assistant for Mac users who want quick automation via chat. ClawdAgent is an autonomous AI operations center — a full-stack, multi-agent, self-evolving system with enterprise-grade security, crypto trading, social publishing, 4 AI providers, 3 agent protocols, and a complete React dashboard. If OpenClaw is a smart personal assistant, ClawdAgent is an AI department.
Features
Core Intelligence — 57 Modules
Multi-Model AI — Claude 4.6 (Anthropic), 400+ models via OpenRouter, local Ollama models, Claude Code CLI (free with Max)
Smart Model Router — Picks the best model per task (complexity, cost, budget) with 5 provider modes
Extended Thinking — Up to 32K thinking tokens for complex reasoning
Streaming Responses — Real-time token streaming across all platforms
Intent Classification — 45+ intents with multilingual support (Hebrew, Arabic, CJK)
Intelligence Bridge — Central nervous system connecting 9 subsystems (scoring, memory, governance, cost, routing, observability, goals, safety, feedback)
Agent Factory — Spawns new specialized agents on demand (Meta Agent)
Crew Orchestrator — Multi-agent teams that collaborate on complex tasks
Autonomous Goal Engine — Self-initiated 30/60/90-day goals with KPIs and milestones
Auto-Upgrade — Checks for and applies skill/prompt/config upgrades automatically
Advanced Capabilities
Browser Automation — Headless Playwright (works on servers without GUI)
Server Management — Multi-server SSH, Docker ops, health monitoring, auto-discovery
Desktop Control — AI vision + mouse/keyboard automation with safety bounds
Mobile Automation — Android via ADB/Appium with pre-built app recipes
Crypto Trading — 5 built-in strategies (DCA, Scalping, Swing, Day Trading, Custom) + full TA engine, risk manager, portfolio tracker
Content Creation — AI video, images, music via Kie.ai (60+ models)
Social Publishing — 9 platforms via Blotato (Twitter, Instagram, TikTok, LinkedIn, YouTube, Facebook, Threads, BlueSky, Pinterest)
Email — Gmail API + SMTP dual integration
Voice — Text-to-Speech + Speech-to-Text via ElevenLabs + AI phone calls via Twilio + OpenAI Realtime
Calendar — Google Calendar integration
Plugin System — Manifest-based extensibility (tools, behaviors, prompts, config)
MCP Support — Model Context Protocol for external tools (9 servers)
YAML Config — Hot-reloadable configuration with 11+ feature toggles
Observability & Cost Intelligence
Timeline Events — 9+ event types tracking every agent action, tool call, and evolution step
Tool Heatmaps — Hourly usage patterns and success rates per tool
Error Clustering — Automatic error categorization and severity tracking
Cost Tracker — Per-model, per-action cost tracking with daily budgets
ROI Analysis — Agent return-on-investment per task type
Budget Forecasting — Token burn prediction and smart provider routing
System Snapshots — Dashboard data snapshots for trend analysis
Agent Interoperability Protocols
ClawdAgent speaks the industry-standard agent protocols, enabling seamless communication with any compliant AI agent:
| Protocol | Standard | Status | Endpoints |
|----------|----------|--------|-----------|
| MCP (Model Context Protocol) | Anthropic | Deep integration | 9 MCP servers, JSON-RPC 2.0 |
| A2A (Agent-to-Agent) | Google / Linux Foundation | Full support | Agent Card, Tasks, SSE streaming |
| ACP (Agent Communication Protocol) | IBM BeeAI / Linux Foundation | Full support | REST runs, agent descriptor |
| Tool Use | Anthropic Claude | Native | 35 integrated tools |
A2A Endpoints
GET /.well-known/agent.json — Public Agent Card (no auth)
POST /a2a — JSON-RPC 2.0 (tasks/send, tasks/get, tasks/cancel)
POST /a2a/stream — SSE streaming (real-time task updates)
GET /a2a/tasks/:id — Get task status
GET /a2a/tasks/:id/subscribe — SSE subscribe to task events
POST /a2a/tasks/:id/cancel — Cancel running task
ACP Endpoints
GET /acp/agent — Agent descriptor
POST /acp/runs — Create run (start processing)
GET /acp/runs/:id — Get run status
POST /acp/runs/:id/input — Continue conversation
POST /acp/runs/:id/cancel — Cancel run
Quick Start
One-Command Install
git clone https://github.com/liortesta/ClawdAgent.git
cd ClawdAgent
bash install.sh
The interactive installer handles everything: prerequisites check, dependencies, environment configuration, security key generation, and build.
Prerequisites
Node.js 18+
PostgreSQL 15+
Redis 7+ (optional — works without it, queues disabled)
pnpm 8+ (auto-installed if missing)
Manual Setup
# 1. Clone
git clone https://github.com/liortesta/ClawdAgent.git
cd ClawdAgent
# 2. Install dependencies
pnpm install && cd web && npm install && cd ..
# 3. Configure
cp .env.example .env
# Edit .env — at minimum set: DATABASE_URL, JWT_SECRET, ENCRYPTION_KEY
# 4. Generate security keys
JWT_SECRET=$(openssl rand -hex 32)
ENCRYPTION_KEY=$(openssl rand -hex 32)
# 5. Start database (using Docker)
docker compose up -d postgres redis
# 6. Build
pnpm run build && cd web && npm run build && cd ..
# 7. Start
pnpm start
Open http://localhost:3000 to see the dashboard.
Minimal Setup (just AI chat)
Only need a database and one AI provider:
DATABASE_URL=postgresql://user:pass@localhost:5432/clawdagent
JWT_SECRET=<generate with: openssl rand -hex 32>
ENCRYPTION_KEY=<generate with: openssl rand -hex 32>
# Option A: Anthropic API key
ANTHROPIC_API_KEY=sk-ant-your-key-here
# Option B: OpenRouter (400+ models, many free)
OPENROUTER_API_KEY=sk-or-v1-your-key-here
# Option C: Claude Code CLI (FREE with Max subscription)
CLAUDE_CODE_ENABLED=true
# Option D: Ollama (local, free)
OLLAMA_ENABLED=true
OLLAMA_URL=http://localhost:11434
Production with PM2
pnpm run build
pm2 start dist/index.js --name clawdagent --max-memory-restart 8G
pm2 save
Docker (Full Stack)
docker compose up -d
This starts ClawdAgent + PostgreSQL + Redis in one command.
The web dashboard provides full control over ClawdAgent:
| Page | Description |
|------|-------------|
| Dashboard | System overview, stats, quick actions, activity feed |
| Chat | Real-time AI chat with syntax highlighting, code copy, markdown tables, export |
| Agents | View and manage all 20 agents + agent stats |
| Tasks | Create and track tasks with status |
| Cron | Schedule recurring jobs with cron expressions |
| Servers | SSH server management + health monitoring |
| Trading | Crypto trading interface + portfolio |
| Browser View | Visual browser with VNC live streaming + AI actions |
| Facebook | Autonomous Facebook agent — accounts, sessions, action logs |
| LinkedIn | Autonomous LinkedIn agent — engagement, posting, networking |
| Twitter/X | Autonomous Twitter agent — tweets, replies, likes, follows |
| TikTok | Autonomous TikTok agent — likes, comments, follows, scroll |
| Voice Agent | Phone call center — inbound/outbound via Twilio + OpenAI Realtime |
| Mobile Agent | Android device control via Appium — app automation |
| Terminal | Web-based terminal for remote command execution |
| Skills | Browse 67 skills, search and filter |
| Knowledge | Knowledge base explorer + RAG document Q&A |
| Intelligence | Intelligence subsystem visualization + metrics |
| Evolution | Self-evolution engine status + trigger cycles |
| Graph | System relationship graph visualization |
| History | Conversation history browser |
| Logs | System logs viewer with filtering |
| Costs | API cost tracking + budget visualization |
| OpenClaw | OpenClaw integration management |
| Settings | Full configuration UI + model selector |
| Login | JWT authentication with role-based access |
Environment Variables
See .env.example for the complete list. Key variables:
| Variable | Required | Description |
|----------|----------|-------------|
| DATABASE_URL | Yes | PostgreSQL connection string |
| JWT_SECRET | Yes | JWT signing key (32+ chars, generate with openssl rand -hex 32) |
| ENCRYPTION_KEY | Yes | Data encryption key (32+ chars) |
| ANTHROPIC_API_KEY | At least one AI provider | Claude API key |
| OPENROUTER_API_KEY | At least one AI provider | 400+ models including free ones |
| CLAUDE_CODE_ENABLED | At least one AI provider | Use Claude Code CLI (free with Max) |
| OLLAMA_ENABLED | At least one AI provider | Use local Ollama models |
| REDIS_URL | No | Redis (queues disabled without it) |
| TELEGRAM_BOT_TOKEN | No | Telegram bot |
| DISCORD_BOT_TOKEN | No | Discord bot |
| WHATSAPP_ENABLED | No | WhatsApp Web (scan QR from dashboard) |
| FAL_AI_API_KEY | No | AI image/video generation (FLUX, Kling, Wan) |
| KIE_AI_API_KEY | No | AI content generation (60+ models) |
| BLOTATO_API_KEY | No | Social media publishing (9 platforms) |
| ELEVENLABS_API_KEY | No | Text-to-Speech + voice cloning |
| BRAVE_API_KEY | No | Web search |
| TWILIO_ACCOUNT_SID | No | Twilio voice calls (SID) |
| TWILIO_AUTH_TOKEN | No | Twilio voice calls (auth) |
| TWILIO_PHONE_NUMBER | No | Twilio outbound phone number |
| OPENAI_API_KEY | No | OpenAI Realtime API for voice AI |
| OPENCLAW_GATEWAY_TOKEN | No | OpenClaw remote agent bridge |
Security
ClawdAgent implements Zero Trust AI Architecture — no component trusts another:
Input → Content Guard sanitizes all input (20+ injection patterns)
Processing → Social engineering detection (15 patterns, auto-block on high severity)
Memory → SHA-256 checksum on store, verify on retrieve, quarantine on tamper
Audit → Tamper-evident hash chain on every operation
LinkedIn Agent — Fully autonomous LinkedIn engagement: like posts, comment with AI-generated text, follow users, send connection requests, publish posts. Cookie-based auth with stealth browser. 1,464 lines of automation code
Twitter/X Agent — Autonomous Twitter engagement: tweet, reply, retweet, like, follow. Cookie-based auth with anti-detection. Full dashboard page with account management, action logs, and configuration
TikTok Agent — Autonomous TikTok engagement: like, comment, follow, scroll, share. Stealth browser with human-like delays. Dashboard page with real-time status
Mobile Agent — Autonomous Android app automation via Appium: control TikTok, Twitter, Facebook natively on real devices. SSH tunnel to Appium server, pre-built app recipes
Voice Agent (Twilio + OpenAI Realtime)
AI Phone Calls — Inbound and outbound phone calls with AI voice powered by Twilio + OpenAI Realtime API
Call Dashboard — Voice Agent page with call history, status monitoring, and configuration
PostgreSQL Call History — All calls logged with transcription, duration, and outcome
Configurable Voice — Choose voice, language, and customize AI instructions per call
Chat Experience Overhaul
Syntax Highlighting — Code blocks now feature full syntax highlighting via highlight.js (40+ languages: JS, TS, Python, Bash, SQL, YAML, and more) with a dark theme optimized for readability
One-Click Code Copy — Every code block has a copy button with visual "Copied" feedback — no more manual selection
Markdown Tables — Pipe-delimited tables (|col|col|) render as styled, responsive HTML tables with alternating row colors
Clipboard Image Paste — Ctrl+V / Cmd+V pastes images directly from clipboard as file attachments — no drag & drop needed
Conversation Export — Export any conversation as Markdown (human-readable) or JSON (structured data) with one click
Message Actions — Hover over any assistant message to copy its text or retry the previous prompt
Mobile-First Sidebar — Conversation sidebar overlays on mobile instead of pushing content, fully responsive
New Dashboard Pages
Browser View — Visual browser with live VNC streaming + AI instruction panel
Facebook Tab — Autonomous Facebook agent management with account, sessions, and logs
LinkedIn Tab — LinkedIn agent dashboard with engagement config
Twitter/X Tab — Twitter agent dashboard with posting and engagement controls
TikTok Tab — TikTok agent dashboard with scroll and engagement automation
Mobile Agent Tab — Android device automation via Appium
Voice Agent Tab — Phone call center with call history and configuration
Terminal Tab — Web-based terminal for remote command execution
Bug Fixes & Stability
Streaming Race Condition — Fixed stale streaming text bleeding between conversations
Thinking Toggle — Fixed dead code in the thinking section collapse/expand logic
File Cleanup Leak — Document uploads now properly clean temp files in all code paths
Pending Response Queue — Added periodic cleanup to prevent unbounded memory growth in WebSocket handler
API Client — Added missing post() method that was breaking mobile agent page builds
Previous (v6.2)
Browser Automation & VNC Streaming
Visual Browser Sessions — Playwright-powered browser with Xvfb + x11vnc + noVNC. Watch your agent browse the web in real-time from the dashboard
ClawdAgent isn't a toy — it's a production system. Here's what people use it for:
| Use Case | How It Works |
|----------|-------------|
| 24/7 DevOps | SSH into servers, restart services, check logs, monitor health, renew SSL certs, alert on issues — all automatic |
| Crypto Trading | 5 strategies (DCA, Scalping, Swing, Day Trading, Custom) across 100+ exchanges with risk management and P&L tracking |
| Social Media Manager | Create AI content (video/image/text) and publish to 9 platforms on schedule with engagement tracking |
| Personal Assistant | Manage calendar, send emails, set reminders, answer questions, remember everything — via WhatsApp or Telegram |
| Code Assistant | Write code, create PRs, review changes, run tests, debug issues — orchestrated by Claude Code |
| Research Analyst | Deep web research, competitor analysis, market reports, saved to knowledge base with RAG |
| Server Fleet Management | Monitor multiple servers, auto-fix common issues, Docker management, health dashboards |
| Content Pipeline | Generate scripts → create videos → make thumbnails → publish everywhere → track performance |
| Mobile Automation | Control Android apps, automate workflows, handle SMS/calls, manage notifications |
| Browser Automation | Fill forms, scrape data, take screenshots, navigate workflows — with VNC streaming so you watch in real-time |
| Facebook Automation | Autonomous agent posts, comments, sends friend requests, joins groups — all with human-like delays and safety limits |
| LinkedIn Automation | Autonomous networking: like posts, comment with AI-generated insights, follow people, send connection requests, publish articles |
| Twitter/X Automation | Autonomous engagement: tweet, reply, retweet, like, follow — with anti-detection and human-like timing |
| TikTok Automation | Autonomous engagement: like, comment, follow, scroll feed, share content — with stealth browser |
| AI Phone Calls | Inbound and outbound calls with AI voice via Twilio + OpenAI Realtime — call center in a box |
| Business Intelligence | Cost tracking, ROI analysis, budget forecasting, agent performance metrics |
| Team Coordination | Multi-agent crews collaborate on complex tasks, with governance and approval gates |
Deployment
Docker (Recommended)
docker compose up -d
Manual (VPS/Cloud)
# Build
pnpm run build
cd web && npm run build && cd ..
# Start with PM2
pm2 start dist/index.js --name clawdagent --max-memory-restart 8G
pm2 save
Requirements for Production
Set NODE_ENV=production
Use strong JWT_SECRET and ENCRYPTION_KEY (generate with openssl rand -hex 32)
Set up PostgreSQL with proper credentials
Configure Redis for full queue support
Set CRON_TIMEZONE to your timezone
Set BIND_HOST=127.0.0.1 behind a reverse proxy (nginx)
Copyright 2024-2026 Lior Ben Shimon (TestaMind)
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Apache 2.0 protects both users and contributors with explicit patent grants and trademark rules, while remaining fully open-source and business-friendly.
