Install Hermes CLI and connect ScriptEngine.

Outcome

What you will configure

Hermes Agent can use any provider that implements the OpenAI-compatible /v1/chat/completions interface. ScriptEngine provides that interface at a single base URL, so the setup requires only an endpoint, an API key and a model ID.

Hermes queries the authenticated ScriptEngine /v1/models endpoint during setup. Choose an ID from that result rather than copying an old model name from another tutorial.

Step 1

Check the requirements

Before installing Hermes, prepare:

• A supported system: Windows 10/11, Apple-silicon macOS, Linux or Windows through WSL2.
• A ScriptEngine account with available credits.
• The full ScriptEngine API key shown after provisioning. It begins with sk-se-.
• A current model ID from the ScriptEngine model list.

Hermes’ official installer handles Python, Node.js and its other managed dependencies. Linux requires Git, curl and xz-utils; see the official installation guide for distro-specific details.

Step 2

Install Hermes CLI

Choose your operating system. These are the current commands published by Nous Research.

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

source ~/.bashrc  # Use ~/.zshrc when your shell is zsh

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

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

Step 3

Verify the installation

Run Hermes’ built-in diagnostic before adding the provider:

hermes doctor

If the shell reports hermes: command not found, start a new terminal first. On Linux or macOS, reload ~/.bashrc or ~/.zshrc and confirm that ~/.local/bin is on your PATH.

Step 4

Connect Hermes to ScriptEngine

Run the full provider wizard outside an active Hermes chat:

hermes model

Then enter the following values:

Hermes has changed the parenthetical wording beside Custom endpoint between releases. It may mention entering a URL manually or self-hosted / VLLM endpoints; both labels lead to the custom base-URL flow used here.

Hermes saves the provider, base URL and model in config.yaml, while the current installer stores the secret separately in ~/.hermes/.env (or the native Windows Hermes home). The settings survive restarts.

Alternative

Configure it manually

Use this fallback when the interactive wizard cannot run. Edit ~/.hermes/config.yaml and ~/.hermes/.env on Linux, macOS or WSL2. Native Windows uses the equivalent files under %LOCALAPPDATA%\hermes.

HERMES_CUSTOM_SCRIPTENGINE_ORG_API_KEY=YOUR_SCRIPTENGINE_API_KEY

model:
  provider: custom
  default: YOUR_MODEL_ID
  base_url: https://scriptengine.org/v1
  api_key: ${HERMES_CUSTOM_SCRIPTENGINE_ORG_API_KEY}
  api_mode: chat_completions

Replace both placeholders locally. Restrict access to the .env file and never commit it or publish a screenshot containing the key.

For current Hermes versions, use model.base_url for a custom provider. Do not rely on the legacy OPENAI_BASE_URL environment-variable route for this setup.

Step 5

Test the connection

Start a one-shot Hermes chat:

hermes chat -q "Reply with exactly: ScriptEngine connected"

A normal model response containing ScriptEngine connected confirms that Hermes loaded the custom provider, authenticated the key and reached the selected model.

Optional: test the API key directly

This request checks the ScriptEngine key and model-list endpoint independently of Hermes:

export SCRIPTENGINE_API_KEY="YOUR_SCRIPTENGINE_API_KEY"
curl https://scriptengine.org/v1/models \
  -H "Authorization: Bearer $SCRIPTENGINE_API_KEY"

$env:SCRIPTENGINE_API_KEY = "YOUR_SCRIPTENGINE_API_KEY"
curl.exe https://scriptengine.org/v1/models `
  -H "Authorization: Bearer $env:SCRIPTENGINE_API_KEY"

A successful JSON model list confirms that the key and base service are reachable. Remove the temporary environment variable or close that terminal when finished.

During a chat

Switch ScriptEngine models

Once the custom provider exists, type /model inside a Hermes session to choose among configured providers and models:

/model

Choose the saved ScriptEngine provider, then select a discovered model. The terminal command hermes model performs provider setup; the in-chat command only switches among providers and models that are already configured.

Fix common errors

Troubleshooting

hermes: command not found

Open a new terminal. On Linux or macOS, reload your shell profile and verify that ~/.local/bin is on PATH. Then run hermes doctor.

HTTP 401 or “invalid API key”

Re-run hermes model and replace the key with the complete active sk-se-… value. Check for spaces, quotes copied into the value, or a rotated/revoked key.

HTTP 404 or endpoint not found

Confirm the base URL is exactly https://scriptengine.org/v1. A duplicated /v1/v1 or an appended /chat/completions in the base-URL field will route incorrectly.

Model not found

Use the exact model ID—not its display label. Check the ScriptEngine Models list or call /v1/models with the API key, then update the provider through hermes model.

Hermes still uses the old provider

Exit the active chat, run hermes model again, and start a fresh session. Inside a session, use /model and choose the saved ScriptEngine provider explicitly.

Verification

Official sources

Hermes is developed by Nous Research; ScriptEngine is a separate custom API provider. Version-sensitive installation and configuration details were checked against these official sources:

• Hermes installation guide
• Hermes custom-provider guide
• Hermes CLI command reference
• Hermes platform-support matrix