Jovethra

Jovethra / Hermes Agent

Run Hermes with a budget you can explain.

Hermes can use a custom OpenAI-compatible endpoint. Keep the API key in its environment file, use a bounded model configuration, and measure the full agent workflow rather than only the visible answer.

01 / Setup

Connect Hermes in two minutes.

Use the endpoint below in your Hermes configuration. The API key remains a server-side secret and must never be committed to a repository.

Example
model:
  provider: custom
  base_url: https://api.jovethra.xyz/v1
  api_key: ${JOVETHRA_GPT_KEY}
  default: gpt-5.6-sol

# ~/.hermes/.env
JOVETHRA_GPT_KEY=your_jovethra_key

02 / Agent usage

One agent turn is not one token.

Count context and tool results as input, the visible answer and model decisions as output, and repeated system or tool context as cache. GPT and Claude keys are separate; input, output, cache, rate, and concurrency limits remain enforced.

03 / Choose a boundary

Start with the smallest plan that contains the workflow.

Starter · €9.99 / month

20M input · 1.25M output · 30M cache

Pro · €19.99 / month

45M input · 2.5M output · 60M cache

Power · €49.99 / month

110M input · 6.25M output · 150M cache