Concepts
This section helps you understand how Silicon Worker works. After reading it, you'll have a clearer picture of what happens behind the scenes while it completes your tasks, and how to make it more controllable and more attuned to you.
How Silicon Worker works
When you give it a task, it doesn't just reply with text. Like a worker, it plans, acts, checks its own work, and pushes forward — stopping to ask you when needed. A few core concepts make this possible:
- How the agent works: how it plans step by step, calls tools, checks results, and advances a task.
- Skills: packaging "the standard way to do a class of task" into reusable capabilities.
- Experts & Teams: dedicated roles for different kinds of work, with multiple roles collaborating on complex tasks.
- Memory: remembering your preferences, processes, and project context across sessions — getting to know you over time.
- Tools & risk levels: every action has a risk level, sensitive actions ask for your consent first, and your data stays local.
Start here
Suggested reading order
Read these in order to build a complete mental model.
- How the agent works: understand how it thinks and acts.
- Skills: teach it the standard way to do a class of task.
- Experts & Teams: set up roles and collaboration for different work.
- Memory: let it remember you and your work.
- Tools & risk levels: understand its safety boundaries.
Next / Related
- Not set up yet? See Install and Connect your first model.
- Want to dive in? See Quickstart.
- Curious which tools it has? See Built-in tools reference.