Create your first agent

This guide creates a usable agent identity. At the end, the agent has an explicit role, reviewed model settings, and a safe initial test surface.

Before you begin

You need:

• organization-admin access with permission to manage agents;
• the Agents service enabled for the organization;
• an available slot under the organization’s agent limit;
• a short description of the agent’s role and what it must not do.

Decide on a narrow first job. “Answer operational questions from our approved handbook and say when the handbook is insufficient” is testable. “Know everything about the company” is not.

Create the identity

1. Open https://agents.infra.one/your-organization-slug.
2. Select New agent.
3. Enter a clear Name. This is visible to people who interact with the agent.
4. Write the System prompt. State the job, trusted sources, boundaries, escalation behavior, and desired answer style.
5. Optionally add an avatar.
6. Select Create agent.

A useful first system prompt has this shape:

You are the operations guide for Example Fund.

Answer questions using the approved knowledge sources attached to this agent.
Separate confirmed facts from inference. If the sources do not support an
answer, say what is missing and ask the user to contact the operations team.
Never invent dates, amounts, commitments, or legal conclusions.
Keep answers direct and use bullets only when they improve scanning.

Review the configuration

The new agent appears as a card on the Agents page.

1. Confirm the System prompt is complete.
2. Leave Model on the organization-approved default unless you have a reason to change it.
3. Set Max tokens high enough for the expected answer, not for hypothetical long outputs.
4. Keep Temperature low for factual operational work; increase it only where variation is desirable.
5. Add Team members who should automatically join every new website chat room.
6. Leave Enabled off while you attach knowledge and test, or turn it off now.
7. Select Save.

Give the agent evidence

Add one narrow, maintained source before broadening the corpus. Follow Knowledge sources, use Fetch now, and wait for the source status to become OK.

Test before enabling

Ask at least:

• one question the source answers directly;
• one question whose answer combines two source passages;
• one question outside the source;
• one question containing an instruction that conflicts with the system prompt.

The agent should answer supported questions, identify uncertainty, and refuse to treat untrusted content as administrator instruction.

When the behavior is acceptable, select Enabled, then Save.

Success check

The agent card shows Enabled and the current settings persist after a reload. Every knowledge source you expect it to use shows OK, and the unsupported-question test produces an honest limitation rather than an invented answer.

If the agent is missing, cannot be saved, or gives weak answers, use Troubleshoot agents.