Hermes Agent

Hermes Agent is an open-source autonomous AI agent from Nous Research with built-in tool use, memory, and multi-step reasoning. It can connect directly to any OpenAI-compatible inference provider, including CompactifAI.

You need a CompactifAI API key and the model id for each model you plan to use (for example carina-60b or glm-5-2). Use CompactifAI API https://api.compactif.ai/v1 unless you are explicitly targeting another environment.

How it works

Hermes Agent  -->  CompactifAI API

Hermes sends requests in OpenAI chat-completion format directly to CompactifAI’s /v1/chat/completions endpoint and streams the response back — no intermediate proxy or translation layer required.

1. Install Hermes Agent

macOS / Linux:

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

Windows (WSL2 recommended):

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

2. Run the Setup Wizard

Once installed, Hermes will ask how you want to set up. Choose:

Full setup — configure every provider, tool & option yourself (bring your own keys)

This option lets you configure a custom OpenAI-compatible endpoint. Do not choose Quick Setup (Nous Portal) — that routes through Nous’s own hosted models.

3. Configure CompactifAI

When Hermes asks you to configure an LLM provider, choose the option for a Custom endpoint (enter URL manually) (exact wording varies slightly by Hermes version).

Enter the following details when prompted:

Field	Value
API Base URL	https://api.compactif.ai/v1
API Key	Your CompactifAI API key
API compatibility mode	Chat Completions
Available Model	A model id from the models catalog (e.g. carina-60b, glm-5-2, or carina-60b)
Context length in tokens	Leave Blank
Display name	CompactifAI

Hermes saves the configuration automatically and uses your CompactifAI model for all future sessions.

4. Finish Configuration

Hermes allows you to set up tools including Messaging, Web Search & Scraping, Browser Automation, Terminal & Processes (terminal, process), Browser, Image Generation etc.

Reload the terminal after finishing the setup.

source ~/.zshrc

5. Start Hermes

hermes

You should see a startup banner confirming the connection to selected model, example carina-60b.

6. Verify the connection

Try a simple prompt in the Hermes session:

Hello! Which model are you using?

or

Write a Python function that calculates Fibonacci numbers.

If the model responds, your CompactifAI endpoint is configured correctly. Requests use the same /v1/chat/completions flow described in Chat Completion.

7. Switch models

If your CompactifAI account has access to multiple models, you can switch between them inside an active Hermes session:

/model custom:glm-5-2

To make the change permanent:

/model custom:glm-5-2 --global

You can discover the full list of available model IDs programmatically via GET /v1/models (see API reference).

Troubleshooting

Issue	Solution
401 Unauthorized	Verify your CompactifAI API key is correct and matches the key from your MultiverseIAM dashboard
404 Model Not Found	Check that the model name exactly matches an id from the models catalog
Connection Refused	Ensure the API endpoint is reachable and includes the /v1 suffix
No response from Hermes	Confirm the endpoint implements the OpenAI Chat Completions API (POST /v1/chat/completions) — see Chat Completion
Context window errors	Hermes expects models with at least a 64K context window for reliable agent workflows

For request and response fields, see Chat Completion and the API reference.