How to Install Hermes Agent and Connect It to AtmoRouter

To use Hermes Agent with AtmoRouter, install Hermes, choose its custom endpoint option, and enter https://atmorouter.dev/v1 with an AtmoRouter API key and an exact model ID. You do not need a different API account for every model family.

This guide covers the terminal workflow. Commands were checked against the official documentation on October 19, 2026; installer options and menu labels can change between releases.

Install Hermes Agent

For Linux, macOS or WSL2, the official installation guide provides this terminal installer:

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

Hermes also documents a native Windows installer. Run its PowerShell command in PowerShell, rather than pasting the Bash command into a Windows terminal:

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

These commands download and execute the official installer. Review it first if that is required by your environment. Reopen your terminal afterward so newly installed commands are available. For a graphical installation on Windows or macOS, use the desktop download linked from the same official guide.

Prepare your AtmoRouter account

Create an account, sign in, and open API Keys. Create a separate key for Hermes so you can revoke its access without changing other applications. Keep the full key private.

Open the model catalogue and copy the complete model ID, including its family prefix. Check context length, input/output prices, and suitability for tool use. A model appearing in the catalogue does not mean it will support every Hermes feature.

Check your account balance before testing a paid model. API usage is separate from installing the Hermes application.

Configure the custom endpoint

In a terminal outside an active Hermes conversation, run:

bashhermes model

Select the custom endpoint option and supply:

• Base URL: https://atmorouter.dev/v1
• API key: the key from your AtmoRouter dashboard
• Model: the exact ID copied from the catalogue
• API mode, if asked: Chat Completions

Use the API base, not the full /chat/completions request path. Hermes documents custom endpoint setup in its provider guide. Its current quickstart requires at least 64,000 tokens of model context; choose a model that meets that requirement rather than entering an invented context limit.

Check one conversation before adding tools

Start Hermes:

bashhermes

Begin with a short prompt such as “Reply with one sentence explaining what an API gateway does.” Confirm that Hermes shows the model you selected and produces an answer. Then open Usage logs to check the model ID, status, timing and cost of the request.

For a coding workflow, continue in a disposable project and ask Hermes to list the project files before approving an edit. This checks more than text generation: the model and client must also cooperate on tool calls.

If the first request fails

• Authentication error: first confirm which Hermes profile is active. Each profile has its own configuration and credentials; saving a key in another profile does not update the running one. Configure the AtmoRouter custom endpoint and full ar- key inside the intended profile, then restart the Hermes process or app for that profile so it loads the saved settings. Never paste the key preview, surrounding quotes, or the text Bearer into a field that asks only for an API key. A revoked key or a key for another service will not authenticate here.
• Model not found: compare the complete model ID against the current catalogue.
• Route not found: remove a duplicated /v1 or a full request path from the base URL.
• Tool errors after a successful chat: test a model appropriate for agent workflows and check the client's selected transport.

Keep error details, timestamps and model IDs when asking for support, but remove keys from screenshots and logs. More endpoint details are in the quickstart and error reference.

Separate key validity from profile setup

The official Hermes profiles guide explains how profile-specific settings and credentials are separated. Check the active profile in your Hermes interface before changing its provider settings; a name assigned to a key in AtmoRouter does not select a Hermes profile.

For an optional key-only check, set ATMOROUTER_API_KEY privately in your terminal environment, then run:

shnpx @atmorouter/[email protected] check-key

Do not put the key in command arguments, screenshots, or support messages. This check uses an empty request body, selects no model, and does not run or bill inference. A successful check confirms that the gateway accepts that environment value; it does not prove that the running Hermes profile uses the same value. Restart the Hermes process or app for that profile after saving its correct settings. Repeatedly retrying an unchanged 401 request will not fix its credentials.

Does the AtmoRouter helper install Hermes?

No. Install Hermes with its own installer and configure its endpoint as described above. The AtmoRouter helper supports a documented set of clients; it is not a universal installer for every AI agent.

For another terminal coding workflow, see OpenCode with AtmoRouter. If you already use OpenRouter, follow the migration checklist.