> ## Documentation Index > Fetch the complete documentation index at: https://docs.dgrid.ai/llms.txt > Use this file to discover all available pages before exploring further. # Use Hermes Agent with DGrid > Configure a custom provider in Hermes Agent and use DGrid models through Chat Completions. > [!NOTE] > > Use a DGrid Model API Key for model requests. A Management API Key is an administrative credential and cannot be used for this tutorial. Check that your key has available credits and access to the selected model. To use DGrid models in Hermes Agent, add a custom API provider and enter your DGrid API key and endpoint. This guide follows the first-time setup wizard and uses `openai/gpt-5.6-sol` as the example model. ## Before you start * A [DGrid account](https://dgrid.ai/) with available credits. * A computer with terminal access. Installation commands for macOS, Linux, WSL2, and native Windows are below. ## Step 1: Create a DGrid API key 1. Sign in to the [DGrid Console](https://dgrid.ai/) and open **Credits → Keys**, or go to the [API Keys page](https://dgrid.ai/api-keys). 2. Select **Create New Key**, give it a name such as `Hermes`, and create it. 3. Copy the key. You will paste it into the Hermes **API key** prompt. Do not include it in screenshots or public documents. ## Step 2: Install Hermes and open setup On macOS, Linux, or WSL2, run: ```bash theme={null} curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash ``` On native Windows, run in PowerShell: ```powershell theme={null} iex (irm https://hermes-agent.nousresearch.com/install.ps1) ``` If you see `prerequisites ok` or `Cloning ... hermes-agent.git`, installation is still running. Wait for the setup wizard. At **How would you like to set up Hermes?**, use the arrow keys to move to **Full setup — configure every provider, tool & option yourself (bring your own keys)**, then press Enter or Space. If installation finishes without opening the wizard, run `hermes model` in your terminal and continue at Step 3. ## Step 3: Choose a custom API provider In **Select provider**, find **Custom (direct API)** near the bottom of the list and select it. If you entered through `hermes model`, choose **Custom endpoint (enter URL manually)**. ![Custom direct API entry in the Hermes provider menu](https://mintcdn.com/dgridai/QQLVyRrVjCz0-l8E/images/integrations/hermes-agent/custom-provider-redacted.png?fit=max\&auto=format\&n=QQLVyRrVjCz0-l8E\&q=85\&s=4514d377cb2b9aa506bd2c867f8d4b67) Custom direct API entry in the Hermes provider menu ## Step 4: Enter the DGrid endpoint and model Follow the wizard prompts in this order: | Wizard prompt | Enter or select | | - | - | | **API base URL** | `https://api.dgrid.ai/v1` | | **API key** | Your DGrid API key from Step 1 | | **Select API compatibility mode** | **Chat Completions**, option `2` in the current wizard | | **Select model** | Full model ID `openai/gpt-5.6-sol` | | **Context length in tokens** | Submit a blank value for automatic detection | | **Display name** | `DGrid` | The base URL stops at `/v1`; do not append `/chat/completions`. The `[optional]` text beside the generic **API key** prompt does not apply when connecting with a DGrid key. After you enter the key, Hermes loads the models visible to it. Find `openai/gpt-5.6-sol` on **Select model** and enter that row's number, or type the full ID. The screenshot shows `96`, but your list number may differ. Use the complete ID to identify the model. If it is absent, choose another available DGrid model that supports Chat Completions. ![The example model and model-selection prompt in Hermes](https://mintcdn.com/dgridai/QQLVyRrVjCz0-l8E/images/integrations/hermes-agent/model-selection-redacted.png?fit=max\&auto=format\&n=QQLVyRrVjCz0-l8E\&q=85\&s=3aa5ee1749a644b9c7e5d9f2dd9b3f67) The example model and model-selection prompt in Hermes When Hermes reports that the model and custom provider were saved, configuration is complete. ## Step 5: Start Hermes and use the DGrid model Run: ```bash theme={null} hermes ``` On first launch, **Help improve Hermes?** may ask for a usage-data preference. Choose your preference; **No thanks** also continues to chat. Enter a message, for example: ```text theme={null} Reply with exactly: DGrid connection successful ``` Once Hermes replies, you can keep chatting. To check that the request went through DGrid, open the DGrid request history around that time and look for the model ID with a **Success** status. The images below show an example reply and request record. ![Chatting with a DGrid model in Hermes](https://mintcdn.com/dgridai/QQLVyRrVjCz0-l8E/images/integrations/hermes-agent/chat-success-redacted-en.png?fit=max\&auto=format\&n=QQLVyRrVjCz0-l8E\&q=85\&s=14257ec4db475941f7c5eeafaf71b804) Chatting with a DGrid model in Hermes ![The model request in the DGrid Console](https://mintcdn.com/dgridai/QQLVyRrVjCz0-l8E/images/integrations/hermes-agent/dgrid-request-success-redacted.png?fit=max\&auto=format\&n=QQLVyRrVjCz0-l8E\&q=85\&s=7f44d7b26671bfee73c25ceddfb0d4db) The model request in the DGrid Console ## Common questions * **The wizard did not open automatically:** run `hermes model`, select the custom endpoint, and enter the values from Step 4. * **Enter inserts a literal `^M`:** press **Control+J** to submit the current text prompt. * **You want to check or change the model:** run `hermes status` in the terminal to see the current Model and Provider; run `hermes model` to select a model again. * **Chat replied, but `Auxiliary title generation failed: Request timed out` appeared:** this message concerns automatic conversation-title generation and does not undo the chat reply. See the [Hermes installation guide](https://hermes-agent.nousresearch.com/docs/getting-started/installation/) and [provider documentation](https://hermes-agent.nousresearch.com/docs/integrations/providers/) for more details. ## Related resources * [Integration tutorials overview](/ai-gateway/integrations.md) * [AI Gateway quickstart](/ai-gateway/overview.md#quickstart) This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.