AI Use Cases

Web.Casa’s AI assistant isn’t a chatbot bolted onto a panel — it’s a co-pilot wired into every subsystem. Here are real scenarios showing how it replaces multi-step manual workflows with a single natural-language request.

Create a Reverse Proxy in One Sentence

Without AI: Open the site creation form, fill in domain, select type, enter upstream address, configure headers, save, wait for certificate issuance.

With AI:

“Create a reverse proxy for api.example.com pointing to localhost:3000 with HTTPS”

The AI calls create_host, configures the upstream, and Caddy provisions the TLS certificate automatically. You review the confirmation dialog and approve — done.

Deploy a Docker App from a Description

Without AI: Write a docker-compose.yml by hand, manage environment variables, create volumes and networks, then start the stack.

With AI:

“Deploy a WordPress site with MySQL and Redis caching”

The AI generates a complete Compose file with correct image versions, environment variables, volume mounts, and network configuration. After your approval, it deploys the stack and reports the access URL.

Diagnose a 502 Error

Without AI: SSH into the server, check Caddy logs, check upstream container status, inspect port bindings, review recent deployments, correlate timestamps.

With AI:

“My site blog.example.com is returning 502, help me investigate”

The AI runs get_host to check the proxy configuration, docker_ps to verify the upstream container is running, docker_logs to scan for errors, and get_system_metrics to check resource usage. It then explains the root cause — for example, the container crashed due to an OOM kill — and suggests a fix.

Manage Your Server from Your IDE

With 44 MCP tools exposed via the Model Context Protocol, your AI-powered IDE becomes a server management interface:

In Cursor / Windsurf / Claude Code:

“Show me all running containers and their resource usage”

The IDE calls docker_ps and get_system_metrics through MCP, displaying results inline. No browser tab, no SSH — just your editor.

More MCP workflows:

• “Deploy the latest commit of my-app” → triggers deploy_project
• “What’s in the build log for the last failed deploy?” → calls get_build_log
• “Create a PostgreSQL database called analytics” → invokes database creation tools
• “Back up everything to S3 right now” → triggers trigger_backup

AI-Powered Code Review

Without AI: Manually review code before deployment, check for hardcoded secrets, missing error handling, SQL injection vectors.

With AI:

“Review the code in my latest commit for security issues”

The AI calls review_code to analyze your changes, checking for:

• Hardcoded credentials or API keys
• SQL injection and XSS vulnerabilities
• Missing input validation
• Performance anti-patterns

It returns a structured report with severity levels and suggested fixes — before the code reaches production.

Generate a Dockerfile from Project Context

Without AI: Analyze the project structure, identify the language/framework, determine the correct base image, write multi-stage builds, configure build arguments.

With AI:

“Generate a Dockerfile for this Next.js project with standalone output”

The AI calls generate_dockerfile, inspects the project’s package.json, framework configuration, and directory structure, then produces an optimized multi-stage Dockerfile with the correct Node.js version, build steps, and production configuration.

Automated System Health Check

Without AI: Manually check disk space, review container status, verify SSL certificates, scan logs for errors across multiple services.

With AI:

“Run a system health inspection”

The AI calls run_inspection, which checks disk usage, memory pressure, container health, SSL certificate expiration, and configuration validity. It returns an AI-generated summary with prioritized recommendations — all in one command.

Self-Healing on Alert

Without AI: Get paged at 3 AM, SSH in, diagnose the alert, take manual action.

With AI: When a monitoring alert fires (e.g., “disk usage > 90%”), the SelfHealEngine automatically:

1. Analyzes the alert context
2. Identifies the cause (Docker image cache bloat, log rotation failure, etc.)
3. Suggests a fix (docker_prune or adjust log retention)

You review and approve — or let it auto-remediate based on your configured policy.

One-Command Full Deployment

Without AI: Create a project, configure Git, detect framework, set up build pipeline, create reverse proxy, enable HTTPS — at least 6 steps.

With AI:

“Deploy https://github.com/my-org/my-app to myapp.example.com”

The AI calls auto_deploy: clones the repo, detects the framework, triggers the build, creates the reverse proxy host, and provisions HTTPS — all in one command.