beginner 10 minutes
Build your first natural-language agent
Ukkin turns a plain-English description into an editable agent. This guide walks through creating one, reviewing its steps, and refining it — the core authoring loop.
- 1
Describe the automation
In Ukkin, describe what you want in natural language, e.g. "watch my wishlist for price drops and tell me when one changes."
- 2
Review the extracted steps
Ukkin interprets your intent with an LLM and shows the workflow it plans to run. Read the steps to confirm they match what you meant.
- 3
Set conditions
Add device-aware conditions so the agent runs at the right time — on Wi-Fi, while charging, or on a schedule.
- 4
Refine by chatting
If a step is off, describe the change and Ukkin edits the agent. Iterate until it does exactly what you want.
Starting from a built-in agent
- ▸Ukkin ships with agents for Instagram mention watching, email triage, and price tracking.
- ▸Cloning and editing one is often faster than describing a similar agent from scratch.
- ▸All agents share the RepetitiveTaskAgent base, so behavior is consistent.
More guides in the index, or read the architecture to see how agents perceive and act.