Configure Hermes Agent with the interactive wizard

Requirements and key groups

Hermes is an independent agent and requires an external-use MonkeyAI key. In key management, choose Claude Plus（精品）, Claude Max（外接版） or a suitable official direct group. For GPT, choose Codex Pro（外接版） or external-capable Lite. Do not use Claude Max（仅限CC） or Codex Pro（仅限Codex）. Lite has no free Fast tier and lists no image models. Review live pricing.

Installation

The official installation guide supports native Windows, macOS including Apple Silicon, and Linux; WSL is optional. Download the official script, review it, then execute it. Reopen the terminal afterward and run hermes --help.

Windows

Invoke-WebRequest https://hermes-agent.nousresearch.com/install.ps1 -OutFile hermes-install.ps1
Get-Content .\hermes-install.ps1
# Execute after reviewing the script
& .\hermes-install.ps1

macOS

curl -fsSL https://hermes-agent.nousresearch.com/install.sh -o hermes-install.sh
less hermes-install.sh
# Execute after reviewing the script
bash hermes-install.sh

Linux

curl -fsSL https://hermes-agent.nousresearch.com/install.sh -o hermes-install.sh
less hermes-install.sh
# Execute after reviewing the script
bash hermes-install.sh

Run hermes model

Use the confirmed interactive model setup entry:

hermes model

Choose a provider with a customizable endpoint or a custom API in the wizard. Enter the following values when prompted; exact labels vary by release:

Setting	Claude path	Codex path
API protocol	Anthropic Messages	OpenAI Responses
Base	https://monkeytheme.com	https://monkeytheme.com/v1
Model ID	claude-opus-4-8	gpt-6.1-sol
Key	Enter your own key privately	Enter your own key privately

Do not append /v1/messages to the Claude base. Confirm that the Codex path supports Responses; a custom OpenAI option meaning only Chat Completions is insufficient. If the required protocol is unavailable, stop that path, check the release or use a supported Claude path. Do not enter a MonkeyAI key into an official OAuth login flow.

Save and launch

Review the wizard's saved provider and model, then run hermes. Restart existing sessions. Configure under the same user and environment that runs Hermes on Windows, WSL, macOS or Linux. Use the interactive wizard to save settings; no manual configuration file is needed.

Verification and recovery

After reviewing pricing, you can ask a short question in an empty directory. Check the key for 401, external-use group for 403, protocol and URL for 404, and exact ID for model_not_found. Diagnose tool failures separately from text request failures: text access does not establish compatibility with every tool, image operation or long-running task. Run hermes model again to restore the previous provider, and retain redacted errors for troubleshooting. See image generation for image capabilities.