Understanding Orquesta's 4 Execution Modes for AI Workflows

Originally published at orquesta.live/blog/understanding-orquestas-4-execution-modes-2026-06-05

Orquesta offers a unique approach to AI-driven workflows with its four distinct execution modes: Auto, SSH, Agent, and Batuta. These modes allow teams to harness the power of AI in a flexible and secure manner, ensuring that prompts are executed optimally based on the task at hand. Understanding when and how to use each mode can significantly enhance your workflow efficiency and collaboration. Let’s delve into each mode and the scenarios they best suit.

Auto Mode: Let the AI Decide

The Auto mode is designed for those times when you want to leverage AI’s decision-making prowess to select the best execution path. When a prompt is submitted in Auto mode, Orquesta delegates the decision to the AI, which evaluates the nature of the task and chooses the most suitable execution method among SSH, Agent, and Batuta.

When to Use Auto Mode:

• Uncertain about the best execution path: If you're unsure whether a task requires a simple SSH command or a more complex autonomous execution, Auto mode can help.
• Maximize efficiency: Letting the AI choose can save time, especially when handling varied or unknown workloads.
• Focus on strategy: Auto mode is ideal when you want to focus on high-level planning, trusting the AI to optimize execution.

SSH Mode: Quick and Direct

SSH mode is the go-to for executing straightforward commands directly on a server. This mode bypasses any AI decision-making, allowing for precise, manual control.

When to Use SSH Mode:

• Simple, clear commands: Best for tasks that require executing a single, well-defined command.
• Low overhead: Use SSH when you need minimal setup and quick execution.
• Direct control: Ideal when you need to maintain complete control over the command execution.

Example Use Case

For instance, if you need to restart a service or check logs on a remote server, SSH mode enables you to execute the necessary commands without any additional overhead or complexity.

ssh user@server 'sudo systemctl restart myservice'

Agent Mode: Full Claude CLI Capabilities

Agent mode takes advantage of the full capabilities of the Claude CLI, allowing Orquesta to execute complex tasks that involve multiple steps, branching logic, or data processing.

When to Use Agent Mode:

• Complex tasks: Best for executing detailed scripts and processes that require more than single-line commands.
• Integration with Claude: If your workflow relies heavily on Claude’s specific functionalities, this mode is optimal.
• Real-time monitoring: Ideal for tasks that benefit from live output streams and detailed monitoring.

Example Use Case

Consider a scenario where you need to deploy a new feature across a microservices architecture. Using Agent mode, you can automate the deployment process, ensuring each service is updated correctly, with real-time feedback.

- task: deploy
  services:
    - name: auth
    - name: payment
  branch: feature-update

Batuta Mode: Autonomous Execution Loop

Batuta mode empowers the AI to operate autonomously, executing tasks based on a Think > Act > Observe > Repeat loop. This mode is particularly powerful for ongoing processes or tasks that require adaptive execution.

When to Use Batuta Mode:

• Long-running processes: Ideal for tasks that require continuous monitoring and adaptation.
• Complex decision-making: Use Batuta when the task involves dynamic environments where decisions need to be made based on real-time data.
• Autonomous workflows: Perfect for scenarios where human intervention is minimal, and the AI can optimize the process iteratively.

Example Use Case

Automating the management of cloud resources based on usage patterns is a prime candidate for Batuta. The AI can monitor resource utilization, scale services up or down, and adjust configurations autonomously.

- task: resource-management
  strategy: adaptive
  criteria:
    - cpu: >70%
    - memory: >80%

Decision Framework for Teams

Choosing the right execution mode depends on your specific use case, team expertise, and the complexity of the tasks at hand:

• Evaluate Task Complexity: Simplistic tasks are best handled by SSH, while multifaceted operations benefit from Agent or Batuta.
• Resource Considerations: Consider the computational and time resources available.
• Desired Control Level: Determine if you need direct control over each step or if you can entrust the AI with autonomous decisions.

By understanding these execution modes, your team can better align their workflows with operational goals, ensuring efficiency and precision in AI-driven tasks. Each mode has its unique strengths, making Orquesta a versatile tool in the modern developer’s toolkit.