> ## Documentation Index > Fetch the complete documentation index at: https://docs.orbitagents.xyz/llms.txt > Use this file to discover all available pages before exploring further. # Build an agent from a goal > Describe what you want done and Agent Builder turns it into a working agent in minutes. When no ready-made [Orbie](/concepts/orbies) fits, **Agent Builder** lets you create your own — you describe the goal in plain English, and it turns that into a working agent in minutes. No prompt-engineering, no technical setup. ## Steps 1. Open **Agent Builder** in your dashboard. 2. Describe the job in plain language — what you want the agent to do, and how you'll know it did it well. For example: *"Every morning, find new sign-ups from yesterday and send me a one-paragraph summary of who they are."* 3. Give it any context it needs (your rules, your tone, what to avoid). 4. Review what it proposes, then create it. Your new agent lands in your company like any other — on the [org chart](/use/organize-org-chart), reading your [decisions](/use/update-decision), filing to your [Vault](/concepts/vault). ## Writing a good goal The clearer the goal, the better the agent. A few tips: * **Say what "done" looks like.** "Summarize" is vague; "send me a 5-bullet summary by email" is clear. * **Name the trigger.** One-off? Daily? When something happens? * **Give the guardrails.** What should it never do? Any rule it must follow? You don't have to get it perfect — you can refine the agent afterward, and it'll follow any [decisions](/concepts/decisions) you've already set. ## Builder vs. Orbie — which to use * **Use an [Orbie](/use/launch-orbie)** when a ready-made agent already does roughly what you need. Fastest path. * **Use Agent Builder** when your job is specific to your business and no ready-made one fits. Either way, you end up with a real teammate in your AI company that shares your memory and reports into your dashboard. Make your new agent run on its own.