Hermes

Configure OpenGateway as a named Hermes provider without changing your default model, then switch sessions to DeepSeek UltraFast or GLM Flash UltraFast.

Copy this to your agent

Paste it into any AI agent. It sets up everything and saves it as the opengateway-hermes skill.

Set it up by hand#

1. Get an API key#

OpenGateway → API Keys → Create Key. Export it in the shell that starts Hermes:

Bash

export OPENGATEWAY_API_KEY="YOUR_OPENGATEWAY_API_KEY"

Add the export to ~/.zshrc or ~/.bashrc if you want it in new terminals. Keep the actual key out of config.yaml and other files that may be committed.

2. Add the provider#

Merge this entry into the model: section of ~/.hermes/config.yaml. Preserve the existing default and other settings under model:

YAML

model:
  providers:
    opengateway:
      base_url: https://apis.opengateway.ai/v1
      key_env: OPENGATEWAY_API_KEY

The opengateway mapping key is the provider name. key_env names the environment variable that contains the API key, so the secret stays outside the YAML file.

3. Switch and verify#

Start a Hermes session after saving the configuration. Inside the session, /model opens the picker; you can choose the configured custom provider and a model. To switch directly to the recommended model, use the documented named-provider form:

/model custom:opengateway:deepseek/deepseek-v4.1-flash-ultrafast

Send "Reply with OK". A response confirms the connection. Usage appears on the dashboard.

Models and limits#

The recommended default model for OpenGateway is deepseek/deepseek-v4.1-flash-ultrafast; z-ai/glm-5.3-flash-ultrafast is an alternative. This guide does not change Hermes's existing global default. See the model catalog for active model IDs and /models for current pricing.

UltraFast models have a per-API-key, per-model concurrency cap. Requests over the cap return 429 with code concurrency_limit_exceeded and a Retry-After header; reduce in-flight requests or contact support to request a higher limit. For both models above, a request with more than 256K total input tokens has double the input-token price for the whole request; cache reads and output tokens are not surcharged. See UltraFast for details.