Managing infrastructure across multiple cloud providers presents a unique challenge. Each platform has its own CLI, APIs, and operational patterns. What if you could interact with all your cloud infrastructure through a single, intelligent interface that understands context, coordinates complex operations, and delegates tasks to specialist agents?
That's exactly what I built with Sky Agent. A multi-cloud orchestration platform that combines Strand's multi-agent capabilities with Claude Code SDK's coding capabilities to create a unified cloud operations experience. This system goes beyond simple command forwarding. It's an intelligent coordinator that analyses tasks, delegates to specialised agents, and manages complex cross-cloud workflows through natural conversation.
The Architecture
Sky Agent (Coordinator) - The intelligent entry point positioned above all cloud operations:
- Analyses incoming requests to identify required cloud providers
- Routes tasks to appropriate specialist agents
- Coordinates multi-cloud operations spanning multiple providers
- Manages cross-cloud resource dependencies and workflows
Specialist Agents:
- AWS Agent - Amazon Web Services operations using AWS Strand's "use_aws" tool
- Azure Agent - Microsoft Azure operations via Azure CLI wrapper
- GCP Agent - Google Cloud Platform management through gcloud wrapper
- Coding Agent - Software development tasks leveraging Claude Code SDK
- Atlassian Agent - Jira and Confluence operations via MCP integration
MCP Servers:
- GitHub
- Atlassian
Sky Agent in Action
The CLI provides an interactive experience for terminal-based workflows:
The web interface through Open WebUI offers a familiar chat experience:
Architectural Decisions
Why AWS Strands Swarm for Coordination?
AWS Strands offers multiple multi-agent orchestration patterns, Workflow and Graph architectures alongside Swarm. After evaluating all three, Swarm emerged as the optimal choice for Sky Agent's dynamic coordination needs.
Why Swarm over Workflow or Graph?
Swarm enables the coordinator to analyse tasks and delegate to specialists on-the-fly, without predefined paths or rigid workflows. Whether handling a simple single-cloud query or orchestrating complex multi-cloud operations, the same architecture adapts seamlessly.
swarm = Swarm(
[sky_agent, aws_agent, azure_agent, gcp_agent, coding_agent, atlassian_agent],
entry_point=sky_agent,
max_handoffs=20,
repetitive_handoff_detection_window=8
)
Why Dedicated Cloud Wrappers?
Rather than forcing uniform patterns, each cloud provider gets a focused wrapper that embraces native CLI tools and authentication mechanisms. This design respects each provider's optimal patterns while maintaining a consistent interface for the coordinator.
Why Containers?
The entire system runs in containers, enabling deployment anywhere - your local machine, AWS, Azure, GCP, or any combination.
What This Enables
Unified Operations: One interface for all cloud providers, eliminating the need to remember provider-specific commands and syntax.
Coordinated Workflows: Complex operations spanning multiple clouds become simple natural language requests.
Development Integration: Code generation, fixes, and Infrastructure as Code workflows integrated alongside cloud operations.
Automated Documentation: Project management integration ensures all operations are tracked and documented automatically.
Conclusion
Sky Agent demonstrates that managing multiple cloud providers doesn't require multiple tools and contexts. By combining intelligent coordination with specialist agents and surgical precision development workflows, it provides a unified interface across your entire cloud infrastructure.
The architecture offers an elevated perspective, coordinating specialist agents who have a deep understanding of their domains. Whether managing one cloud or three, the system adapts to your infrastructure while maintaining consistent operations.
Top comments (0)