AI agents · Types
What is an autonomous agent?
The short answer
It is given a goal, not steps. It plans, acts, checks the result and adjusts, and it can keep working for minutes, hours or days.
It is the one form of autonomous agent.
| What starts it | A goal, from a person or an event |
|---|---|
| Who decides the next step | The agent plans its own steps |
When should you use an autonomous agent?
Use it when the path cannot be written down in advance: research, investigation, diagnosis.
When should you not?
Not when the steps are known. Small errors compound over a long run, and a pipeline will be cheaper and more reliable.
What does it need to run safely?
It needs a limit on steps, time and spend, a contained place to work, points at which it reports or asks, and a full record of what it did.
An example
Given "find out why last night's settlement run failed", it reads the logs, tests its explanations and writes up the cause with the evidence.
What else is it called?
Long-horizon agent, deep research agent.
How do you build one on Katonic?
- New agent, Answer & assist, then Research report. Start blank. The sections are Brief, Sources, Depth & budget, Report, Guardrails, Advanced.
- Brief. Name the agent, pick the Owner team, and write the Research question: state the question and what a good answer looks like. Add Scope & constraints to keep it on target.
- Sources. Turn on the sources it may search.
- Depth & budget. Choose Focused or Wide, set the follow-depth and the source cap, then a time budget and a cost budget.
- Report. Set the sections of the report.
- Test. Choose Start research run. The run is asynchronous and shows four phases: planning, searching sources, drafting the report, verifying citations.
- Deploy, then Publish.
Questions people ask
What starts an autonomous agent?
A goal, from a person or an event.
Who decides its next step?
The agent plans its own steps.