Best Practices

Alex offers several modes of work. The right approach saves time, request units, and leads to better results. This page describes recommended workflows from simple tasks to deep analysis.

Complexity Levels

Simple Tasks — Direct Query

For quick actions, just write what you need:

Restart the Minecraft server

Check how much RAM the server is using

Alex executes directly. No planning needed.

Medium Tasks — Context in One Message

For multi-step tasks, give Alex all information at once. The more context, the better the result.

Bad:

Install WordPress.

(later)

Now set up SSL.

(later)

And add email too.

Good:

Install WordPress on domain example.com, set up a Let’s Encrypt SSL certificate, and create an email mailbox [email protected].

Alex plans the approach and executes everything in one pass — fewer request units, better results.

What to Include in Context

• Which server — “on VPS for domain example.com”
• Desired outcome — “I want a working WordPress with HTTPS”
• Constraints — “use PHP 8.2, nginx instead of Apache”
• Priority — “backup first, then make changes”

Complex Problems — Planning Mode

For deeper problems (debugging, outage analysis, optimization), activate Planning Mode — icon in the chat header. In this mode, Alex does not execute actions, only analyzes and prepares a plan.

Why Use Planning Mode

• Saves request units — Alex analyzes without running commands
• Better analysis — Alex has time to review logs, configuration, structure
• Clear plan — You get a structured list of steps to execute
• Approval before action — You see exactly what Alex will do before it happens

How To

1. Activate Planning Mode — icon in the chat header
2. Describe the problem in detail:

Server example.com returns 502 Bad Gateway. Nginx is running but the backend
on port 8080 is not responding. Check logs, nginx configuration, PHP-FPM status,
and propose a fix plan.

3. Alex analyzes — reviews logs, configuration, processes and prepares a plan
4. Review the plan — you’ll see a list of steps with reasoning
5. Deactivate Planning Mode and tell Alex to execute the plan

Tip: In Planning Mode, Alex uses sub-agents (explore, error-detective, debugger) for deeper analysis. Each sub-agent specializes in a specific aspect of the problem.

Architectural Tasks — Planning + Sub-agents + Implementation

For the most complex tasks (new infrastructure, migration, large refactoring), use the full workflow:

Step 1 — Planning Mode

Activate Planning Mode and describe the goal:

Propose a migration from Apache to nginx. Currently we have:
- 3 WordPress sites
- 1 Laravel application
- PHP 8.2 via FPM
- MySQL database
I need a zero-downtime migration.

Alex creates a detailed plan with steps, risks, and recommendations.

Step 2 — Delegation to Sub-agents

After approving the plan, Alex automatically delegates tasks to specialized sub-agents:

• explore — maps the current Apache configuration
• backend-architect — proposes nginx configuration
• devops — performs the actual migration
• security — reviews the resulting configuration

Each sub-agent works independently on its part. Alex coordinates their work and consolidates results.

Step 3 — Step-by-Step Implementation

Alex executes the plan step by step:

1. Creates a backup of the current configuration
2. Implements the first step (e.g., installs nginx)
3. Verifies the step worked
4. Moves to the next step
5. After each significant step, you can review the state and decide whether to continue

Important: Alex always asks for confirmation before destructive operations. You can say “stop” at any time and Alex will halt.

Quick Reference — When to Use What

Tips for Effective Communication

Be Specific

Less effective:

Something is not working.

Better:

Server minecraft.example.com has been unresponsive since 14:00. The console shows an OutOfMemoryError.

Provide Context

I have a Minecraft server with 50 players, 32 GB RAM, Spigot 1.20.4.
After installing WorldEdit plugin the server is lagging.
Check TPS and suggest optimization.

Split Large Tasks

If you need to make 10 different changes, split them into 2-3 messages instead of one massive request. Alex maintains context better and you can check results along the way.

Leverage Memory

Tell Alex important information once and it will remember:

For future reference: our production server is a VPS on coresynth.io,
domain example.com, we use Docker, nginx, and PHP 8.2.

In the next conversation, you won’t need to repeat this context.

Use Interactive Questions

Alex can ask for your decisions via interactive cards. If you want Alex to ask before every important choice, say:

Before each important change, ask for my approval.

Next Steps

• What is Alex — Feature overview
• Superpowers — Guided workflows
• Interactive Questions — How Alex asks for decisions
• Models & Limits — AI models and message limits
• FAQ — Frequently asked questions

Need help? Open a support ticket  or ask Alex in your panel.