Hermes Agent

Hermes Agent, the terminal agent from Nous Research, has no built-in ozeye provider and does not need one: its custom endpoint takes an OpenAI-compatible base URL.

Quick setup

hermes config set OZEYE_API_KEY oz_your_key_here
hermes model      # choose "Custom endpoint", base URL https://ozeye.ai/v1, model glm-5.3

Manual configuration

# ~/.hermes/.env
OZEYE_API_KEY=oz_your_key_here

# ~/.hermes/config.yaml
model:
  provider: custom
  base_url: https://ozeye.ai/v1
  key_env: OZEYE_API_KEY
  default: glm-5.3

fallback_providers:
  - provider: custom
    base_url: https://ozeye.ai/v1
    key_env: OZEYE_API_KEY
    model: kimi-k2.6

auxiliary:
  title:
    provider: custom
    base_url: https://ozeye.ai/v1
    key_env: OZEYE_API_KEY
    model: glm-5.3-flash
  compression:
    provider: custom
    base_url: https://ozeye.ai/v1
    key_env: OZEYE_API_KEY
    model: glm-5.3-flash

The fallback_providers and auxiliary blocks are optional: a fallback model keeps a session alive if the primary is unavailable, and cheap auxiliary models (titles, context compression) keep long sessions inexpensive. Launch with hermes or hermes --tui.

Model ids

Use the ids from the models page as they are: glm-5.3, kimi-k2.6, glm-5.3-flash. Pick models with at least 64k context for agent sessions. Every chat model on the list qualifies except codestral-latest and qwen2.5-vl-72b-instruct, which have 32k.

Troubleshooting