Skip to main content
Hermes Agent supports custom OpenAI-compatible endpoints, so it can connect directly to Mercury 2.5 without a proxy.
Mercury 2.5 supports parallel tool calls, allowing Hermes to run independent tools in the same turn.

​
Prerequisites

​
Setup

1

Install Hermes Agent

Follow the Hermes installation guide, or install it on macOS, Linux, or WSL2:
Open a new terminal after installation so the hermes command is available.
2

Add Inception as a custom endpoint

Open Hermes’s model picker:
Select Custom endpoint (enter URL manually) and enter:If Mercury 2.5 does not appear during model discovery, enter mercury-2.5 manually. Hermes stores the key in its user-only .env file and saves Mercury 2.5 as the default model.
3

Start Hermes

Launch an interactive session:
The startup message should show mercury-2.5 with the custom provider.

​
Verify the integration

Run a one-shot prompt:
Hermes should respond with Mercury 2.5 is connected.
  • Hermes cannot find the endpoint: Confirm the base URL ends in /v1 and the compatibility mode is Chat Completions.
  • Mercury 2.5 is not listed: Enter mercury-2.5 manually in the custom endpoint flow.
  • The wrong model is active: Run hermes model again and select the saved Inception provider.
Hermes custom-provider documentation