About 5 min read
Create and run your first agent
Creating an agent, choosing a sensible starting model, and verifying health before you rely on it for real work.
Create an agent
- Start from a template when available; it encodes sensible defaults.
- Name agents for the role they perform so teammates can discover them quickly.
Pick a model policy
Match model capability to task risk: cheaper models for classification, stronger models for multi-step reasoning.
If you use workspace API keys, ensure the provider account has budget alerts.
Verify health
- Use built-in health or ping actions where exposed in your workspace.
- If the agent depends on a gateway or external channel, confirm that dependency is connected before debugging prompts.
Related articles
- What to do right after you sign up — A short checklist after you create an account: confirm email, pick or wait for your workspace, and open the product for the first time.
- Workspaces explained — How a workspace relates to your company account, why you might have more than one, and where your agents and data live.
- Sign in, SSO, and launch links — How sign-in works with the control plane, branded workspace URLs, and secure handoff into your tenant runtime.
- Members, invites, and roles — Inviting colleagues, what new members should expect on first visit, and keeping access aligned with how your team works.
- Billing, plans, and invoices — Where billing is managed, how usage-related charges may appear, and how to get help when something looks wrong on an invoice.
- Help Center home