/

Create your first agent

Pick an agent type, fill in the builder, and use the readiness check to confirm the agent will actually work.

Before you start

Some agents need something connected first. It’s quicker to do this before you build.

Type Connect first
Telecaller A telephony line. See voice calling setup.
Email Agent A mailbox in Relay.
Chat Agent Nothing for the website widget. A Facebook page in Relay for Messenger.
Research Agent An Exa or Tavily key under Research Agent settings.
Marketing Agent A sending mailbox in Relay and consented segments in Nest.
Finance Agent A sending mailbox in Relay. Stripe or Razorpay keys if you want payment links.

Create an agent

  1. Open https://tron.oroe.ai/ and select New Agent.
  2. Choose a type card. The type can’t be changed later.
  3. Under Identity, enter an Agent Name.
  4. Optional: pick a Purpose preset (optional). It pre-fills the name, persona and tone, and switches on a sensible set of tools.
  5. Optional: pick an Industry playbook (optional). It seeds the agent’s settings with an industry’s defaults once, when the agent is created.
  6. Write the Persona / System Prompt: who the agent is, its goal, what it knows and its rules.
  7. Work through the other cards (below), then select Save Agent.

Your agent now appears on the type’s Agents page.

Tip Start from a preset. It’s the fastest way to a working agent, and you can change every value afterwards.

The builder cards

Which cards you see depends on the type.

Card What it’s for
Identity Name, preset, playbook, persona. Telecallers also have Greeting Text (optional).
Voice Telecaller only: TTS Model, Voice, Language, speed and a voice preview.
AI Model Provider, Model and Temperature for an agent on its own key.
Execution Where the model runs: Its own key (legacy) or Shared Pulse connection.
Conversation Telecaller only: Personality / Tone, Max Call Duration (seconds) and an optional Conversation Flow (optional).
Tools What the agent may do. Switch off anything it shouldn’t use. Nothing selected means every tool of the type.
Type settings The type’s own settings, grouped into Identity and goal, Business sources, Tools and channels, Execution, Approvals, Triggers and Budget.
Guardrails Forbidden Topics (comma-separated) and Required Disclosures. These are added to the agent’s instructions.
Status Agent is active. Switch it off to pause the agent.

Run the agent on Pulse

The Execution card shows four checks: Pulse connected, Rollout enabled, Background grant and, once you pick a policy, Tool-ready.

  1. Save the agent first. A binding needs a saved agent.
  2. Under Execution policy, choose a policy.
  3. Optional: under Knowledge, tick up to 10 Pulse collections the agent may quote.
  4. Select Use the shared connection.

To go back, select Back to its own key.

Note Execution policies are created in Pulse by a workspace owner. If the list is empty, see Connecting AI models.

Check readiness

Once an agent is saved, reopen it. A Readiness card shows whether it will actually work, computed from what is configured today.

Each row is a capability, channel or tool. Each row has five columns:

Column Meaning
Available This build implements it.
Connected The thing it talks to is configured.
Permitted This agent is allowed to use it.
Approval A person decides before it happens.
Tested Somebody proved it works. Rarely known.

The card’s badge reads Ready, Not ready or Unknown. When something is blocking, a line starting Blocked: names the one thing to fix first.

Good to know Unknown means nobody could answer the question, for example because Pulse couldn’t be reached. It isn’t the same as broken, and it isn’t the same as working.

You can also see every agent’s readiness at once at the bottom of each settings page, under Agent readiness.

Manage your agents

On a type’s Agents page you can Edit, Duplicate or Delete an agent. Deleting can’t be undone.

Next steps

Was this helpful? Tell us what’s missing with the feedback button in any module, or email support@oroe.ai.