Skip to main content
  1. Add an LLM credential.
  2. Open Agents and select New agent.
  3. Name the outcome or role clearly.
  4. Select the provider credential and one of its available models.
  5. Write system instructions that define the goal, boundaries, evidence requirements, and output style.
  6. Attach only the connections the agent needs.
  7. Create a job with a concrete prompt and run it manually.
  8. Review the complete run history, not only the final answer.
Start with a read-oriented task such as a monitoring brief. Add write actions only after the output is reliable and the external credential is narrowly scoped.