First, one distinction
Hermes Desktop is the Hermes GUI.
Hermes Desktop is Nous Research’s native chat application. It is not a separate agent: the Desktop app and command-line interface share the same Hermes runtime, provider configuration, API keys, sessions, skills and memory.
You configure the ScriptEngine custom provider once. Desktop then uses the same saved provider and model as the CLI, so you can move between interfaces without maintaining two API configurations.
The browser-based hermes dashboard is a different management interface. This tutorial covers the native Desktop app.
Step 1
Prepare the account and key
- Create a ScriptEngine account and add credits.
- Store the full provisioned API key beginning with
sk-se-. - Choose a current model ID returned by ScriptEngine’s authenticated model discovery.
- Use Windows 10/11, Apple-silicon macOS, or a supported Linux installation.
The key belongs in the provider configuration field. Keep it out of prompts, screenshots, repositories and shared shell transcripts.
Step 2
Install Hermes Desktop
Windows 10 or 11
- Open the official Hermes website.
- Choose the Windows Desktop download and run the installer.
- Launch Hermes Desktop. On first launch, it can provision the Hermes runtime and managed dependencies under
%LOCALAPPDATA%\hermes.
The Desktop installer is the simplest Windows route and does not require administrator access for the managed per-user installation.
Apple-silicon macOS
- Open the official Hermes website.
- Download the macOS Desktop installer and open it.
- Complete the normal macOS application installation, then launch Hermes Desktop.
The official Hermes platform matrix lists Apple-silicon macOS as Tier 1 and Intel macOS as unsupported.
Linux
Install Hermes CLI first, then launch the Desktop build through the shared installation:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
source ~/.bashrc
hermes desktop
The official installer notes that Desktop native modules require a compiler toolchain. On Debian or Ubuntu, install build-essential if the build requests it.
Step 3
Complete first launch
Let Desktop finish its managed Hermes installation. If onboarding offers provider setup before you are ready, choose the option to configure a provider later and open the main application.
Hermes Desktop should now have a chat composer, project sidebar and Settings area. The model picker sits beside the composer; Settings → Model controls the profile-wide default after a provider has been configured.
Run hermes desktop. Desktop reuses the existing config, keys, sessions and skills instead of creating a second agent.
Step 4
Add ScriptEngine in Desktop
Open Settings → Providers → Custom Endpoints, then use the Add Endpoint form.
| Desktop field | Enter |
|---|---|
| Name | ScriptEngine |
| Provider ID | scriptengine |
| Endpoint URL | https://scriptengine.org/v1 |
| Default Model | Choose after the authenticated model test |
| Context | Leave blank / Auto unless a verified limit is published |
| API Key | Your complete private sk-se-… key |
- Enable Use for new chats.
- Enable Discover models.
- Select Test. Desktop calls the authenticated
/v1/modelsendpoint. - Choose one of the discovered model IDs as the default, then select Save.
https://scriptengine.org/v1 is an inference endpoint. It belongs under Providers → Custom Endpoints, not Settings → Gateway → Remote gateway.
If your Desktop build does not show Custom Endpoints
Some earlier Desktop releases expose less of this flow. Update Desktop, or configure the same shared runtime once from a terminal:
hermes model
Choose the option whose label begins Custom endpoint, enter https://scriptengine.org/v1 and your key, choose Chat Completions, select a discovered model, leave context blank, and use ScriptEngine as the display name. The parenthetical wording beside Custom endpoint varies between Hermes releases. Relaunch Desktop afterward.
Step 5
Return to Hermes Desktop
- Quit and reopen Desktop if it was running while you changed the provider from a terminal.
- Start a new chat. The saved ScriptEngine endpoint should be available to the current profile.
- Open the composer’s model picker to confirm the selected custom model.
- If another model is selected, choose the configured custom provider and the ScriptEngine model.
The composer model picker changes the current chat or Desktop-local selection. Use for new chats sets the saved custom endpoint as the default for new Desktop conversations.
Alternative
Manual configuration fallback
If neither Desktop setup nor the provider wizard can run, close Desktop and edit the shared Hermes configuration:
- Windows:
%LOCALAPPDATA%\hermes\config.yamland%LOCALAPPDATA%\hermes\.env - macOS or Linux:
~/.hermes/config.yamland~/.hermes/.env
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 the placeholder only in your local file, save it, and relaunch Desktop. Do not append /chat/completions to the base URL.
Step 6
Test the first Desktop chat
Create a new chat and send:
Reply with exactly: ScriptEngine connected
A normal model response containing ScriptEngine connected confirms that Desktop loaded the shared custom provider, authenticated your key and reached the selected model.
If you prefer an independent check, follow the direct /v1/models key test in the Hermes CLI tutorial.
Fix common errors
Troubleshooting
- The custom model does not appear in Desktop
- Run the endpoint’s Test action again. If your release lacks the Custom Endpoints section, quit Desktop, run
hermes model, finish setup, and relaunch the app. Both interfaces must use the same Hermes home/profile. - Windows Desktop and CLI show different settings
- Confirm both use
%LOCALAPPDATA%\hermes. A WSL2 installation uses~/.hermesinside Linux and does not automatically share native Windows configuration. - HTTP 401 or authentication failed
- Re-run the provider wizard with the complete active ScriptEngine key. Check whether the key has been rotated, revoked or copied with surrounding whitespace.
- HTTP 404 from the model request
- Set the base URL to exactly
https://scriptengine.org/v1. Remove duplicate/v1segments or request-specific paths. - Desktop opens but no response arrives
- Run
hermes doctorand then the one-shot CLI test from the companion tutorial. If the CLI test succeeds, relaunch Desktop and start a fresh chat.
Verification
Official sources
Hermes Agent and Hermes Desktop are developed by Nous Research. ScriptEngine is a separate custom API provider. This guide was checked against: