Hermes Agent

Hermes Agent supports an OpenAI-compatible Custom endpoint. Prepare a key and model using the shared guide.

1. Install Hermes

macOS, Linux, or WSL2:

curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash

Windows PowerShell:

iex (irm https://hermes-agent.nousresearch.com/install.ps1)

Open a new terminal after installation.

2. Configure the endpoint

Exit any active chat, then run:

hermes model

Choose Custom endpoint (self-hosted / VLLM / etc.):

Use hermes model outside a chat to add endpoints; /model inside a chat switches configured models.

3. Check the saved configuration

On macOS/Linux/WSL2, the file is ~/.hermes/config.yaml. The equivalent Chat Completions configuration is:

model:  provider: custom  default: YOUR_MODEL_ID  base_url: https://api.vnaipro.app/v1  api_key: YOUR_API_KEY  api_mode: chat_completions

Merge into the existing model section, replace both placeholders, and preserve other tool settings. Native Windows installs under %LOCALAPPDATA%\hermes by default; prefer the wizard to target the correct profile.

For a model requiring another protocol, set model.api_mode: codex_responses for Responses or anthropic_messages for Anthropic Messages. Responses keeps the /v1 base URL; Anthropic uses the gateway root https://api.vnaipro.app. The model/group must support that protocol.

4. Test a new session

Run hermes, send a short prompt, and ask it to read a file in a test directory. Check usage history on the website. If requests still use another service, inspect model.provider, model.base_url, and the active profile. Search and other external tools may require separate keys.

Sources: Installation, AI providers, Model configuration.