Skip to main content

Hermes Agent

Hermes Agent is an open-source personal AI agent from Nous Research. It supports custom OpenAI-compatible endpoints, tool use, terminal tasks, and messaging gateways.

Recommended: generate the config from the API Key page​

  1. Sign in to MuskAI and open API Keys.
  2. Click Use Key for the key you want.
  3. Select Hermes Agent.
  4. Click Download config.yaml, then merge its model section into ~/.hermes/config.yaml.

The generated file includes the current API key, MuskAI Base URL, group default model, and the compatible wire protocol.

:::warning Merge instead of replacing If Hermes already has tools, gateways, memory, or other providers configured, merge only the model section. Replacing the whole file removes those settings. :::

Manual configuration​

model:
  provider: custom
  default: "gpt-5.6-sol"
  base_url: "MuskAI Base URL ending in /v1"
  api_key: "your MuskAI API key"
  api_mode: chat_completions

Keep api_mode: chat_completions. Do not append /chat/completions or /responses to base_url.

Why the API mode is pinned​

Hermes can infer the wire protocol from the provider, model, and endpoint. GPT-5.x may otherwise be upgraded to the Responses API. Mixing a Responses SSE stream with a Chat Completions parser can produce:

Provider returned an empty stream with no finish_reason

The generated MuskAI configuration explicitly uses Chat Completions so the request and stream parser stay on the same protocol.

Run and verify​

Follow the current installation instructions in the official Hermes Agent repository, then run:

hermes

Send a small test task and verify that the MuskAI usage page shows a /v1/chat/completions request for the selected API key. If only /v1/models appears, Hermes has only probed the model list and has not sent an inference request yet.

Recent Hermes versions use ~/.hermes/config.yaml as the source of truth for models and endpoints. Do not rely only on the legacy OPENAI_BASE_URL or LLM_MODEL environment variables.