Using Corion AI with Hermes Agent

Hermes Agent is a terminal coding agent that works with any OpenAI-compatible endpoint. Pointing it at Corion takes one interactive setup step.

Prerequisites

Hermes Agent installed
A Corion API key from Dashboard → API keys

Fast setup

hermes model

In the interactive menu:

Select Custom endpoint (sometimes shown as "self-hosted / VLLM / etc.").
Enter the Corion API base URL, ending in /v1:
```
https://api.corion.ai/v1
```
Enter your Corion API key when prompted.

Enter the model identifier exactly as Corion expects. Confirm your key can see it first with curl https://api.corion.ai/v1/models -H "Authorization: Bearer YOUR_KEY".

Model	Identifier
Kimi K3 (1M context, 128K max output)	`kimi-k3`
DeepSeek V4 Pro 0813 (1M context, 375K max output)	`deepseek-v4-pro-0813`
DeepSeek V4 Flash 0731 (1M context, 64K max output)	`deepseek-v4-flash-0731`

Keep API mode on auto-detect — Corion speaks the standard chat-completions protocol.

Then start Hermes:

hermes

Hermes appends the chat-completions path itself — give it the API root (https://api.corion.ai/v1), not https://api.corion.ai/v1/chat/completions.

Manual configuration

Alternatively, edit ~/.hermes/config.yaml:

model:
  provider: custom
  default: kimi-k3
  base_url: https://api.corion.ai/v1
  api_key: your-corion-api-key

Verify

Run a quick prompt and confirm a response streams back:

hermes "Write a two-line launch announcement."

Start with a read-only question ("summarize this directory") and check both the reply and any tool calls before granting write access.

If you get a 401, the key is wrong or incomplete — Corion shows the complete key only once at creation, so generate a fresh one in the dashboard. For other errors, see errors and retries.