6 - Create Agent to Coordinate Skills
- How to build a Joule agent
Prerequisites
- You have completed the previous tutorial in the Joule Studio CodeJam Mission, Create Skill to Create / Update Shipment.
Joule agents represent the next evolution of enterprise automation — intelligent, autonomous systems that plan, reason, and act across multiple tools and systems to achieve complex goals.
While Joule Skills execute predefined, deterministic operations, Joule agents handle multi-step, adaptive workflows, deciding what to do, when, and how based on business context and user intent.
Functioning as orchestration layers, Joule agents combine analytical reasoning with real-time decision-making. They use Large Language Models (LLMs) to interpret user requests, dynamically select relevant tools (including Joule Skills), and generate contextual responses — enabling goal-oriented, conversational automation.
Each Joule Agent is built around four key cognitive abilities:
-
Planning – The agent determines the best sequence of actions to achieve a business goal, orchestrating multiple tools or Joule Skills as needed.
-
Reflection – The agent evaluates its own steps, identifies errors or missing data, and self-corrects to reach the desired outcome.
-
Tool Usage – The agent dynamically invokes SAP Build Actions, Joule Skills, or other APIs to perform operations, retrieve data, or trigger systems.
-
Collaboration – Agents can cooperate with other agents or human users, engaging in dialogue, confirmation, or validation when business logic requires oversight.
While Joule Skills define what actions can be performed (e.g., analyze workloads or simulate optimizations), the Joule Agent defines how these skills are orchestrated in a conversational, goal-driven workflow.
This agent:
-
Combines multiple Joule Skills into an intelligent flow.
-
Makes autonomous decisions based on data and context.
-
Communicates naturally with warehouse supervisors.
In short, this is where your Logistics Agent scenario becomes fully functional.
Joule agents also enable the use of the following tools and capabilities which are not available via skills.
Document Grounding: You can have AI Core index or “embed” a set of documents and use them as context inside your agent.
MCP Servers: You can connect your agent to MCP servers and use their tools inside your agent.
LLM Settings: You can specify which LLMs you want to use, and set whether you want addition pre- and post-processing when the agent is figuring out what to do.













