Create an agent

Before creating

Before opening Create Agent, confirm that:

• You are signed in and invite verified.
• The account has an active subscription.
• The current plan still allows another agent of the selected type.
• If Agent ID binding is required, you have an available Agent ID.
• The account has usable model configuration for runtime creation.
• If you want Hermes Expert Profiles, your plan still has available quota.

These checks prevent agents from being created in a state that cannot run, cannot be billed correctly, or conflicts with identity binding.

Fields

Agent ID
Agent ID is the agent's long-term identity. It powers email, wallet, and future identity features. Some plans require an Agent ID. If you already have active agents, the form also avoids conflicting identity bindings.

Agent name
Required. Use a responsibility-based name such as Community Support, Trading Monitor, or Member Concierge. A good name helps the team understand what the agent owns.

Description
Optional. Use it to describe the audience, channel, or scope, for example "answers project FAQ and event timing questions in the Telegram group."

Agent type
The current create flow supports Hermes and OpenClaw.

• Hermes: for long-running collaboration, specialist tasks, and Expert Profile scenarios.
• OpenClaw: for operational tasks in chat apps, inboxes, email, calendars, and similar workflows.

The page checks your plan and existing agent count before allowing another agent of that type.

Model
The model is the default AI model used by the agent. During runtime creation, Hast passes the model configuration to the runtime. Usage is reflected in Usage and Billing.

Expert Profile
Hermes can create Expert Profiles when the plan allows it. Use them to split one Hermes agent into specialist directions such as research, operations, finance, or technical support. Availability depends on plan capability and remaining quota.

Brain Restore
If you have a .tar.gz brain backup, upload it during creation to restore existing agent data. If not, skip this field.

After submission

After you click Create Agent, Hast will:

1. Save the agent configuration.
2. Derive runtime resources and capabilities from the current plan.
3. Send agent type, model, region, Agent ID capabilities, Expert Profiles, and Brain Restore information to the runtime service.
4. Create the runtime instance.
5. Move the agent to provisioning, then wait for status sync to report running.

If runtime creation fails, the page shows an error. Fix the cause and create the agent again.

Common failures

• Invite required: the account has not passed invite verification.
• Plan required: the account has no active subscription.
• Limit reached: this plan has no remaining slot for the selected agent type.
• Missing model key: the runtime cannot receive model-call capability.
• Agent ID conflict: active agents already use a different identity binding.
• Expert Profile quota reached: Hermes Expert Profile quota is exhausted.
• Deployment failed: the runtime service could not create the instance. Check model configuration, region, and plan state before retrying.