Aura Method (Prompt-Based)

How it works

Aura translates a plain-language prompt into a draft canvas with nodes, edges, tools and variable suggestions. You can accept, refine, or extend the proposal on the canvas.

1. Open Graph Studio Go to the Graph Studio page in your project.

2. Start a new Agent Graph with Aura Choose “Create with Aura” and describe your objective, inputs, data sources, tools, and expected outputs.

   Tip

   Specify relevant collections and tools to improve grounding and structure.

3. Review the draft Copilot generates Start, Agent, Tool, and End nodes, connects edges, and proposes variables.

   Validate assumptions and refine prompts, RAG settings, tools, and variables.

4. Refine and save Edit agent prompts, attach collections, wire variables, and save the graph.

   Check

   You should see a coherent path from Start to End with any branches clearly labeled.

Prompt Tips

• Goal: what the Agent Graph should achieve and for whom
• Inputs: list input variables and formats
• Knowledge: name relevant collections or domains
• Tools: APIs or actions to call
• Success criteria: expected outputs or acceptance checks

Tip

Use variables to avoid re-computing values and to pass context across agents and tools.