Build your own Agents

Build, customize, and orchestrate your AI team to support day-to-day operations.

Welcome to CloudThinker's Multi-Agent System (MAS) platform, where you can create your own team of specialized AI agents to transform your cloud operations. This guide will walk you through how to build, customize, and orchestrate your AI team to handle complex cloud management tasks with unprecedented efficiency.

1

Understanding Agent Orchestration and Multi-Agent Systems (MAS)

CloudThinker Prompting Guide

2

Agent Instruction - Customizing Your Agent prompt

Each CloudThinker agent can be customized through detailed instructions that shape their behavior, expertise, and approach to problem-solving.

How to Configure Agent Instructions:

1. Navigate to the Agent Settings panel (as shown in the screenshots for Alex)

2. Select the "Instructions" tab

3. Define the agent's:

   • Professional background and expertise

   • Core responsibilities

   • Decision-making guidelines

   • Communication style

   • Collaboration protocols

Example Instruction (Alex - Cloud Engineer):

You are a certified AWS Solutions Architect Professional with 8+ years of experience in cloud infrastructure and DevOps. You've successfully implemented cost optimization strategies that saved millions in cloud spending for enterprise clients. Your expertise spans across infrastructure as code, containerization, serverless architectures, and automated deployment pipelines. You're known for your ability to identify cost-saving opportunities while maintaining high performance and security standards.

Guidelines:
- Manage cloud resources and services
- Monitor performance and costs
- Ensure security compliance
- Execute changes with AWS CLI tools

Ensure you leverage #mem #search #alert when you need it.

This comprehensive instruction ensures Alex understands his role, expertise level, and operational guidelines when responding to your requests.

3

Agent Tools - Allow to Agent access to external resources

CloudThinker agents can be equipped with various tools that extend their capabilities and enable them to perform specific actions. As seen in the screenshots, tools are configured in the "Tools & Services" section.

Built-in Tools:

• AWS Script Execution (Read Only/Write Permissions): Allows agents to run AWS CLI commands

• Create Chart: Generate visual representations of cloud data

• Push Alert: Send notifications about critical events

• Create Recommendations: Formulate optimization suggestions

• Search Internet: Gather relevant information from online sources

• Planning: Create structured plans for complex operations

MCP Servers:

Agents can also connect to various Model Context Protocols which support SSE, example:

• kubectl: For Kubernetes cluster management

• pg: For PostgreSQL database management

Configuring Tools:

1. Navigate to the agent's "Tools & Services" tab

2. Check/uncheck tools according to your security requirements and the agent's responsibilities

3. Save Changes to update the agent's capabilities

By carefully selecting which tools each agent has access to, you can maintain proper security boundaries while enabling effective operation.

4

Agent Context - Allow your Agent self-discovery environment (optional)

The Agent Context's ability to enable self-discovery and environmental awareness that helps your agents understand your specific environment and make more relevant recommendations.

Types of Context Information:

• Organization: Company structure, team setup, and priorities

• Environment: Details about your cloud infrastructure, applications, and architecture

• Policies: Security requirements, cost constraints, and operational guidelines

• Historical Data: Previous incidents, common issues, and established solutions

• And more based on your Instruction for Agent

Enable Agent Context:

1. Navigate to the "Agent Context" tab in Agent Settings

2. Enable or disable Agent Context

You can guide the Agent to use context more effectively by updating the Agent Instructions to tell the Agent what information should be discovered and stored in the context.

---

By effectively utilizing CloudThinker's Multi-Agent System, you'll transform your cloud operations from manual, reactive firefighting to an intelligent, proactive partnership between your team and specialized AI agents. Start building your dream team today and experience the future of cloud management.

Tony - Database Engineer

Kai - Kubernetes Engineer

Alex - Cloud Engineer

Anna - General Manager