Create your first agent
An agent uses instructions and tools to act on a task. Begin with one well-defined job, such as summarizing a ticket from a connected source. Keep write actions separate until you know how the agent behaves with test inputs.
Read Agent basics and Add an Agent.
Write instructions that describe the goal, allowed sources, and when the agent should stop or ask for help.
Give it only the tools and files it needs. Check each tool’s access and credentials.
Test the agent with safe records and both normal and unexpected requests.
Choose where to run it, then monitor outcomes.
See Instructions, tools, skills, files, and models for setup details and Agent templates for reusable starting points.
Was this helpful?