Connect the 2,000-Free-Calls API to Hermes Agent and OpenClaw

A step-by-step guide to using ModelScope API Inference as an OpenAI-compatible low-cost endpoint for Hermes Agent and OpenClaw.

This article is for people who already have Hermes Agent or OpenClaw running, but want a lower-cost API path. ModelScope API Inference is useful here because it exposes an OpenAI-compatible endpoint and currently advertises a free daily quota for light workloads.

1. Prepare only two values

In practice you only need the ModelScope token as api_key and the API Inference base_url. The model name is chosen later inside each agent.

api_key: ModelScope Token

base_url: https://api-inference.modelscope.cn/v1

2. Get the ModelScope token

Before using API Inference, bind the Alibaba Cloud account and complete real-name verification. Then open access control and copy either the default token or a newly created one.

3. Connect the free API inside Hermes Agent

Run hermes model, enter More providers, choose Custom endpoint, paste the base_url and api_key, then pick a model from the visible list. If you expect long sessions, set the context length high enough.

hermes model

Context length: 512000

4. Connect the same free API inside OpenClaw

Run openclaw config, choose Local gateway, open the Model section, select Custom Provider, paste the base_url and api_key, then find a valid Model ID from ModelScope and continue the setup. Restart the gateway before opening the TUI.

openclaw config
openclaw gateway restart
openclaw tui

5. Smallest useful test

Do not start with a complex workflow. First send a very small prompt such as “introduce yourself” or “summarize this paragraph”. If that succeeds, the route is alive.

6. Practical reminder

This free route is good for learning, prototypes, and lightweight daily automation, but it is not the right choice for high-concurrency or strict production-SLA workloads.