Requirements vs Workflows

Makinari uses two different engines to execute automation: Requirements and Workflows. Understanding the difference is key to using the platform effectively.

Requirements (AI Goals)

A Requirement is a non-deterministic, objective-based goal handled by AI Agents.

• How it works: You write instructions in plain text (the “Brain”). The CMO agent analyzes it, plans the necessary steps, and delegates Commands to specialized agents.
• Execution: Agents use reasoning to decide which tools to call. If a tool fails, the agent can adapt and try a different approach.
• Use case: Complex tasks requiring creativity, research, or adaptation (e.g., “Research these 50 leads and write personalized outreach emails”).
• Cost: Bounded by a Token Budget.

Workflows (Temporal)

A Workflow is a deterministic, code-based orchestration powered by the Temporal engine.

• How it works: Workflows execute predefined code paths. They are triggered by API calls (e.g., POST /api/workflow/leadGeneration).
• Execution: Workflows follow exact steps. They are highly reliable, can run for days, and can survive server restarts. They often call AI Agents for specific sub-tasks, but the overarching flow is rigid.
• Use case: System processes, syncing large amounts of data, or standard operating procedures (e.g., “Every morning at 8 AM, pull new emails and run the Daily Standup”).
• Cost: Bounded by server execution time.

Summary

• Use Requirements when you want the AI to figure out how to solve a problem.
• Use Workflows when you have a strict, repeatable process that must execute perfectly every time.