Your first open-source agent, in six steps.
Every tool and link here is checked against its official source and re-checked monthly.
- 01
What an agent is
An agent takes a trigger, plans, uses tools, acts and learns from the result. You set the intent and the limits.

- 02
Choose one starter
Personal assistant: Hermes Agent. Visual builder: n8n or Dify. Team of agents: CrewAI. Not sure? Describe your goal in the marketplace's Find my tool box.
- 03
Set up safely
Run it on your own computer first. Pick one model provider (or a local model via Ollama), put the key in a .env file and never share it.
- 04
Your first workflow
Start small: a 9:00 daily briefing or inbox triage that drafts replies. Run it by hand three times before you schedule it.
- 05
Keep a human in the loop
Allow find and draft. Require approval to send. Block delete. Review the action log weekly and know your kill switch.
- 06
Grow from here
Move to a private server only when it runs reliably, add one tool at a time, and try the simulators to test your oversight design.
Open simulators →