Hermes Agent API setup: configuration and troubleshooting

Hermes Agent is an open-source general-purpose agent from Nous Research that runs in a terminal and can connect to chat apps. It supports custom OpenAI-compatible endpoints, so pointing it at FlyRouter is enough.

Updated:

Set up with the wizard

Run the model wizard

Terminal
hermes model

Choose the custom endpoint

Select Custom endpoint (self-hosted / VLLM / etc.) and answer the prompts:

PromptValue
API base URLhttps://api.flyrouter.ai/v1
API keyYour FlyRouter key (sk-...)
Model nameA model ID, for example claude-sonnet-5-5

Choose chat_completions as the API mode. You can leave the context length blank.

Edit the config file

The wizard writes ~/.hermes/config.yaml, which you can also edit by hand. Keep the key in ~/.hermes/.env and reference the variable name from the config:

~/.hermes/config.yaml
providers:
  flyrouter:
    api: https://api.flyrouter.ai/v1
    key_env: FLYROUTER_API_KEY
    transport: chat_completions
model:
  default: claude-sonnet-5-5
  provider: custom:flyrouter
~/.hermes/.env
FLYROUTER_API_KEY=sk-...

Switch models

Use /model inside a session, in the form custom:provider name:model ID:

text
/model custom:flyrouter:gpt-6.1-sol

/model only switches between providers that are already configured. To add one, run hermes model.

Troubleshooting

  • 401: the variable name in ~/.hermes/.env must match key_env. Hermes does not reuse OPENAI_API_KEY for custom endpoints.
  • 404: api must include /v1.
  • Model not found: use the model ID exactly as shown on the pricing page.