Hub.ai

Hub.ai is the first fully AI-managed cloud platform. We provide infrastructure that can be controlled entirely through AI using the Model Context Protocol (MCP), allowing you to provision and manage VMs, containers, and complete infrastructures using natural language through your favorite AI tools like ChatGPT, Claude, or custom agents.

Key features include:

• Full MCP API for AI-driven infrastructure management
• Instant VM and container provisioning
• AI-enhanced web terminals with command generation and explanation features
• Full root SSH access to all resources
• Site hosting with automatic SSL and custom domains
• Resource pools for organizing infrastructure

The Model Context Protocol (MCP) is a standardized JSON-RPC 2.0 interface that allows AI systems to interact with cloud infrastructure. Hub.ai implements MCP to enable:

• Natural language infrastructure provisioning
• AI-driven resource management and scaling
• Automated resource provisioning and configurations
• Integration with any MCP-compatible AI tool

You can connect any AI that supports MCP to Hub.ai and start managing infrastructure through conversation. See our MCP documentation for integration details.

Hub.ai is designed for:

• AI/ML Engineers: Provision and manage infrastructure through AI agents
• Developers: Quickly spin up development environments with natural language
• DevOps Teams: Automate infrastructure management with AI
• Startups: Get enterprise-grade infrastructure without the complexity
• Students & Learners: Learn cloud infrastructure through AI assistance
• Anyone who wants to manage cloud infrastructure without dealing with complex configurations

Hub.ai integrates with AI in three ways:

1. MCP API: Connect any AI tool (ChatGPT, Claude, etc.) to Hub.ai's MCP endpoint. Your AI can then create VMs, manage containers, and handle all infrastructure tasks.
2. AI-Enhanced Web Terminal: Every terminal includes AI features for generating commands and explaining terminal output.
3. Hubert (AI Chat): Built-in chat assistant that can perform operations on your resources and provide system administration help.

Example: Tell your AI "I need a web server with 4GB RAM running nginx" and it will provision the VM, install nginx, and configure everything automatically.

You can sign up or sign in using:

• Email & Password: Create a traditional account
• Google OAuth: Sign in with your Google account
• GitHub OAuth: Sign in with your GitHub account

Visit the Sign In / Sign Up page to get started. All methods provide the same full access to Hub.ai features.

Hub.ai uses a simple credit-based system where 1 credit = $1.00 USD:

• Pay-as-you-go: Credits are consumed per minute while resources are running
• Containers cost ~50% less than equivalent VMs
• Stop anytime: Pausing resources instantly stops billing
• No hidden fees: Transparent pricing for all resources

Example pricing:

• Small VM (2 vCPU, 4GB RAM): ~$0.05/hour
• Production VM (8 vCPU, 16GB RAM): ~$0.25/hour
• GPU-enabled VM: Additional ~$0.35/hour

Use promo code FIRST50 for 50% off your first credit purchase!

Credits can be purchased through the Billing section of your account:

1. Navigate to Account → Billing
2. Click "Top Up Credits"
3. Select a credit package or enter a custom amount
4. Complete payment via Stripe (supports all major cards)
5. Credits are added instantly to your account

Minimum purchase is $10. Enterprise customers can contact us for volume discounts.

No, credits never expire. Once purchased, they remain in your account until used. This allows you to:

• Buy credits in bulk for better rates
• Pause projects without losing credits
• Use credits at your own pace

Monitor your usage in multiple ways:

• Account Dashboard: Real-time credit balance and usage stats
• Billing History: Detailed transaction logs in Billing section
• Resource Page: See per-resource consumption rates
• Email Alerts: Optional notifications for low credit balance

Connecting your AI to Hub.ai is simple:

1. Get an API Key: Generate an MCP API key from Account → API Keys
2. Configure your AI tool: Add Hub.ai as an MCP server with endpoint https://hub.ai/mcp
3. Authenticate: Use your API key as a Bearer token
4. Start building: Your AI can now provision and manage infrastructure

Example configuration for Claude Desktop:

{
  "mcpServers": {
    "hub-ai": {
      "url": "https://hub.ai/mcp",
      "apiKey": "hub_api_your_key_here"
    }
  }
}

Hub.ai provides 42 MCP tools organized into these categories:

VM Management (9 tools): vm_create, vm_update, vm_action, vm_list, vm_get, vm_exec, plus job detail endpoints

Container Management (9 tools): container_create, container_update, container_action, container_list, container_get, container_exec, plus job detail endpoints

File Operations (8 tools): file_read, file_write, file_edit, file_multi_edit, file_ls, file_glob, file_grep, file_grep_raw

Site Management (12 tools): site_create, site_delete, site_domains_update, site_sync, site_rule_add/update/delete, site_list, site_get, plus job detail endpoints

Job Management (5 tools): job_status, job_wait, job_result, job_list_active, job_list_by_resource

Resource Management (4 tools): pool_list, pool_get, template_list, user_limits (includes site limits)

See the full MCP tools documentation for complete details.

Yes! The MCP protocol is just JSON-RPC 2.0, so you can call it directly from any programming language:

curl -X POST https://hub.ai/mcp \
  -H "Authorization: Bearer hub_api_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "vm_list",
      "arguments": {}
    }
  }'

This makes Hub.ai perfect for automation scripts, CI/CD pipelines, and custom integrations.

API keys have granular permissions to ensure security:

• mcp.access: Required for all MCP operations
• Full access: Same permissions as your user account
• Read-only mode: Coming soon for monitoring-only access

Keys are prefixed with hub_api_ for easy identification. You can create multiple keys for different applications and revoke them anytime from the API Keys page.

Virtual Machines (VMs):

• Full operating system with dedicated kernel
• Complete isolation from other VMs
• Can run any OS or kernel version
• Supports desktop environments (Ubuntu Desktop, etc.)
• GPU passthrough available
• Best for: Production workloads, Windows apps, custom kernels

Containers (LXC):

• Lightweight Linux environments
• Share host kernel (more efficient)
• Start in seconds
• ~50% cheaper than equivalent VMs
• Best for: Microservices, development, most web apps

For VMs:

• Ubuntu (22.04 LTS, 24.04 LTS, Desktop versions)
• Debian (11, 12)
• CentOS Stream (8, 9)
• Rocky Linux (8, 9)
• Fedora (latest versions)
• Windows Server (2019, 2022) - Coming soon

For Containers:

• Ubuntu (20.04, 22.04, 24.04)
• Debian (11, 12)
• Alpine Linux (lightweight)
• Rocky Linux
• Custom images - Coming soon

VMs:

• CPU: 1-32 vCPUs
• RAM: 512MB - 64GB
• Disk: 10GB - 1TB
• GPU: Optional (RTX 4090 available)
• Network: Public IPv4/IPv6 optional

Containers:

• CPU: 1-16 vCPUs
• RAM: 128MB - 32GB
• Disk: 1GB - 500GB
• Network: Public IPv4/IPv6 optional

Your account has limits on total resources. View current limits in your account dashboard.

Multiple access methods are available:

• Web Terminal: Full terminal access directly in your browser with AI command generation and text explanation
• SSH: Full root access via SSH (keys managed automatically)
• AI Control: Manage through natural language via MCP
• Remote Desktop: VNC/RDP for desktop OS versions

All instances come with full root access - install anything, configure everything, no restrictions.

Yes, all instance data is persistent:

• While running: All data saved to disk is preserved
• When stopped: Disk state is maintained (no charges for compute)
• When deleted: Data is permanently removed

Important: Always backup critical data externally. We recommend using object storage or external backups for important files.

Yes, instances can be resized:

• VMs: Can upgrade CPU, RAM, and disk (requires brief restart)
• Containers: Can be resized on-the-fly in most cases
• Downgrades: CPU and RAM can be reduced, disk cannot be shrunk

Resize through the web interface or via MCP API. Billing adjusts automatically to new resource levels.

Hubert is the AI Chat Assistant available at /chat that can perform operations on your resources. Access Hubert from any resource's detail page by clicking "AI Assistant" or "Chat". Hubert can:

• Execute Commands: Run commands directly on your VMs and containers
• Install Software: "Install and configure PostgreSQL"
• System Administration: Start/stop services, manage files, check logs
• Diagnostics: "Check disk usage" or "Why is my app using so much memory?"
• Multi-step Tasks: "Set up a LAMP stack with SSL"

Hubert has full context about your resource and can execute administrative tasks through natural language.

The AI-enhanced web terminal at /terminal provides intelligent features for command-line work:

• Full bash shell: Complete Linux terminal with root access
• Command Generation: Describe what you want to do and get the exact command
• Text Explanation: Select any terminal output to get AI explanations
• Error Analysis: Get help understanding error messages
• Copy/paste support: Full clipboard integration
• Session persistence: Terminal sessions survive page refreshes

Yes, AI can execute commands in two ways:

1. Via MCP RemoteBash: Your external AI can run commands through the API
2. Hubert Chat: Hubert can directly execute commands on your resources

This dual approach gives you automation power while maintaining security and control.

Sites is Hub.ai's reverse proxy service that routes web traffic to your instances:

• Automatic domains: Every site gets a *.hub.ai subdomain
• Custom domains: Add your own domains with automatic SSL
• Path-based routing: Route different paths to different instances
• Load balancing: Distribute traffic across multiple instances
• SSL certificates: Automatic Let's Encrypt SSL for all domains
• WebSocket support: Full support for real-time applications

Routing rules map incoming requests to your instances:

1. Create a site in the Sites section
2. Add routing rules with:
   • Path prefix: e.g., /, /api, /blog
   • Target instance: Select from your VMs/containers
   • Target port: Port where your app listens (e.g., 3000, 8080)
3. Save and your site is live!

Example: Route myapp.hub.ai/ → Container "frontend" port 3000, and myapp.hub.ai/api → VM "backend" port 8080

Adding custom domains is simple:

1. Add your domain to a site in the Hub.ai dashboard
2. Create a CNAME record pointing to your network zone's target domain (check your site settings for the specific target)
3. SSL certificates are automatically provisioned
4. Your domain goes live after DNS propagation

Multiple domains can point to the same site, perfect for handling www and non-www versions.

Security is built-in:

• Automatic SSL: All sites get free Let's Encrypt certificates
• Force HTTPS: Optional redirect from HTTP to HTTPS
• Security headers: Automatic security headers for protection
• DDoS protection: Basic protection included
• IP filtering: Coming soon

Resource Pools are containers for organizing your infrastructure:

• Logical grouping: Organize VMs and containers by project, environment, or team
• Isolation: Resources in different pools are logically separated
• Easy management: View and manage related resources together
• AI context: Helps AI understand your infrastructure organization

Example: Create separate pools for "production", "staging", and "development" environments.

Resource Pool limits depend on your account type:

• Free tier: 2 resource pools
• Standard accounts: 5 resource pools
• Pro accounts: 10 resource pools
• Enterprise: Unlimited resource pools

View your current limits in the account dashboard.

Background Jobs handle long-running operations asynchronously:

• VM/Container creation: Track provisioning progress
• Resource updates: Monitor resize operations
• Bulk operations: Handle multiple resource changes

Access jobs via the jobs icon in the header when logged in. Each job shows:

• Current status and progress percentage
• Step-by-step progress updates
• Estimated completion time
• Error messages if something goes wrong

Multiple monitoring options are available:

• Dashboard: Real-time resource status and metrics
• Background Jobs: Track all async operations
• Terminal: Use standard Linux tools (top, htop, etc.)
• AI Monitoring: Use Hubert chat to check system health
• Metrics API: Coming soon for external monitoring tools

Account limits vary by tier:

Standard accounts:

• Max 10 VMs
• Max 20 containers
• Max 50 sites
• Max 5 resource pools
• Max 32 vCPUs per VM
• Max 64GB RAM per VM

Pro accounts:

• Max 25 VMs
• Max 50 containers
• Max 100 sites
• Max 10 resource pools
• Higher per-instance limits

View your specific limits in your account dashboard. Contact sales for enterprise limits.

Yes, limits can be increased by:

• Upgrading your account: Move to Pro or Enterprise tier
• Requesting increases: Contact support for specific limit increases
• Enterprise agreements: Custom limits for large infrastructures

Many limit increases can be approved quickly. Contact us to discuss your needs.

Security is our top priority:

• Infrastructure: Enterprise-grade Proxmox virtualization
• Isolation: Full VM isolation with KVM hypervisor
• Encryption: All data encrypted in transit (TLS 1.3)
• Authentication: Secure JWT tokens with regular rotation
• API Keys: Scoped permissions and easy revocation
• Network: DDoS protection and firewall rules

Your data is your own:

• No access: We cannot access your VM/container data
• Full control: You have root access to all resources
• Data location: Currently US-based infrastructure
• Deletion: Data is immediately purged when resources are deleted
• GDPR: Compliant with data protection regulations

See our Privacy Policy for full details.

Currently, backups are the user's responsibility:

• Manual backups: Use standard tools (rsync, tar, etc.)
• Snapshots: Coming soon for VMs
• Automated backups: On the roadmap for Pro/Enterprise

We recommend implementing your own backup strategy for critical data.

Try these steps:

1. Check resource limits: Ensure you haven't exceeded your quotas
2. View job status: Check the background jobs for error details
3. Verify configuration: Ensure CPU/RAM/disk values are valid
4. Try a different template: Some OS templates may have temporary issues
5. Contact support: If problems persist, contact us with the job ID

Connection issues can have several causes:

• Instance status: Ensure the instance is "Running" not "Stopped"
• Network configuration: Check if public IP is enabled (if needed)
• Firewall rules: Ensure your application ports are open
• Service status: Use web terminal to check if services are running
• Sites routing: Verify routing rules point to correct ports

The web terminal always works for running instances, so start debugging there.

Common MCP connection issues:

1. API Key format: Ensure key starts with hub_api_
2. Authorization header: Use Bearer hub_api_your_key
3. Endpoint URL: Should be https://hub.ai/mcp (not /mcp/v1/message)
4. Content-Type: Must be application/json
5. JSON-RPC format: Ensure proper JSON-RPC 2.0 structure

Test with curl first to isolate AI tool configuration issues.

Optimize performance with these tips:

• Right-size resources: Ensure adequate CPU/RAM for your workload
• Use containers: When possible, containers are more efficient than VMs
• Geographic location: Resources are US-based; consider latency
• Caching: Implement caching in your applications
• Monitoring: Use top/htop to identify bottlenecks

Use Hubert chat or the AI-enhanced terminal features for performance optimization suggestions specific to your setup.