Hermes Deployment

Create an independent Hermes container instance for each Application. By default, use a base image with a fixed release version and retain native functionality; the platform provides purchasing, renewal, persistent storage, and deployment management.

Creation and Purchase

  1. Open Hermes Deployment and create an Application.
  2. Select a duration plan on the purchase page and complete payment.
  3. Return to the configuration page, where you can select a platform chat model from the drop-down list and view the current model price. The platform will create independent API credentials for this instance; you can also disable platform integration and configure it yourself in the native console.
  4. Choose to automatically generate access credentials, or set your own strong password/Token; select A2A and EnvarPay as needed, then confirm to begin deployment. Access credentials and model API credentials are independent of each other.
  5. After the instance is ready, obtain the access address and login credentials from the management page.
  6. Log in to the native console to configure channels. If platform model integration was disabled during deployment, you must also configure the model provider and credentials yourself before chat can be used.

Deployment fees are charged according to the plan duration. For specific USD prices and durations, refer to the purchase page. Model invocation fees are charged separately; purchasing deployment duration does not include model credits. When platform integration is enabled, dedicated API credentials only expose the chat interface corresponding to the selected model, and usage is charged separately from your platform balance. Each Application is purchased, renewed, and managed separately.

Passwords, Credentials, and Subsequent Configuration

Access credentials are configured before the first startup. Automatic generation uses random credentials, custom passwords must contain at least 16 characters, and OpenClaw Tokens must contain at least 32 characters. Drafts do not save passwords; you may need to re-enter them after switching pages. After deployment, only the Application owner and administrators can view access credentials.

Model API credentials for each instance are not reused. Their validity is extended upon deployment renewal and they are revoked when the instance is deleted. If the platform model was not connected during the initial deployment, you can select a platform model on that Application's configuration page and redeploy; the platform will update the default model while preserving conversations and other native settings. Afterwards, modify models and channels in the native console; if switching to a model from another service, first adjust the corresponding interface authorization in platform credential management, or use the corresponding credentials. Normal redeployment does not overwrite native settings. Deployment environments that have not passed release verification will temporarily reject new deployments; please contact support.

Optional Deployment Components

After selecting A2A, the deployment template will start the native Hermes A2A gateway and provide an owner token on the management page. Selecting EnvarPay will additionally install an independent service container and persistent ledger, and requires A2A to be enabled first. EnvarPay initially only has owner access credentials, without a service or payment recipient identity, and will not automatically charge fees or make purchases on your behalf. Neither option automatically registers the instance with Envar; existing instances can enter configuration from the management page, enable them, and redeploy. Enabled components cannot currently be disabled in this wizard.

After A2A is enabled and the instance is ready, the access address card on the management page will display a copyable Agent Card address. Native Hermes A2A uses an independent public address; the Dashboard login address cannot be used as an Agent Card address. If EnvarPay is also enabled, the management page will additionally display the corresponding “Envar Access Address”; when connecting services to Envar, prioritize copying this URL.

Versions and Data

The configuration page displays the currently released image version. New instances use the service's currently released version; existing instances are not automatically replaced when the catalog is updated. Redeployment applies the current service version; please back up first and confirm the upgrade requirements of the version in use.

Each instance provides by default 1 CPU, a 2 GiB memory limit, and 10 GiB of persistent storage. Instance data is retained after restart. Actively deleting a deployment or recycling it after expiration will delete storage; please export any data that needs to be retained in advance.

Renewal and Expiration

Renew on the purchase page of the current Application, and the validity period will be extended according to the platform's existing duration rules. After expiration, access entry is suspended; renewing during the grace period restores access. After the grace period, the instance and data will be recycled. Refer to the expiration time, billing status, and deletion notices on the management page.

Logging In to the Management Interface

On the deployment management page, copy the Dashboard address, username admin, and password, then open it and log in. Login information is provided only to the owner of that Application and platform administrators. Models, skills, and channels are configured in the native Hermes interface.

Frequently Asked Questions

  • Why can the new instance not respond yet? Deployment completion only means that the container is ready. If platform model integration was disabled during deployment, you must select a platform model on the configuration page and redeploy, or configure the model and credentials yourself in the native interface.
  • Are additional plugins or model fallbacks preinstalled? They are not added by default; configure them as needed in the native interface.
  • How do I update an existing instance? The original Application and order are retained, and the current service version is used during redeployment; back up data and check version compatibility before upgrading.
  • What should I do if startup fails? First check deployment logs and events, and configure the model only after confirming that the instance is ready. Do not expose instance passwords and keys in public comments or screenshots.