Skip to main content
Use the low-code Agent builder to create Atlas AI agents that solve business problems and automate workflows. Start from scratch or use a template. Test your agent with the chat preview, then publish it to the Agent library for all CDF project users.

​
Before you start

Complete these steps before building your agent:
  1. Scope your use case — define the specific problem or workflow you want to automate.
  2. Identify an evaluation dataset — gather sample data to test your agent’s performance.
  3. Choose a language model — select the language model that best fits your use case.

​
Build and publish an agent

1

Navigate to Agent builder

In CDF, navigate to Atlas AI > Agent builder.Select + Create agent or select a template to use as a starting point.
2

Configure basic information

Enter a Name that clearly describes the agent’s purpose.Write a Description that explains what problems the agent solves or tasks it automates.Add Sample prompts that show users how to interact with the agent. Make these specific and representative of real use cases.
3

Set up prompting

Select the Language model you identified earlier.Write Instructions that define what you want the agent to accomplish and how it should achieve it. Be clear and specific about:
  • The agent’s role and expertise.
  • The expected output format.
  • Any constraints or limitations.
  • How to handle edge cases.
See Prompts and prompt engineering for more details on writing effective instructions.
4

Configure tools

Add Tools to let the agent access CDF data, perform complex tasks, or interact with other applications.For the data retrieval tools, specify the data model and view to query, and which access scope to use:
  • Inherit scope from the user’s location: the location configured in CDF determines which data to find.
  • Inherit scope based on user’s access rights: find data without filtering on spaces.
  • Follow access scope defined manually: select the spaces to scope the query to.
5

Test and refine

Test the agent using the chat interface. Refine the language model, instructions, and tools as needed.Test with various scenarios, including:
  • Typical use cases
  • Edge cases
  • Error conditions
  • Different types of user input
6

Publish

Select Publish to make the agent available in the Agent library for all CDF project users.
7

Monitor and iterate

Monitor the agent’s performance and make ongoing improvements based on user feedback and usage patterns.