NanoClaw, OpenClaw

NanoClaw and OpenClaw are lightweight Python agent runners. They don’t bundle credential management. Authsome fills that gap. The same patterns work for any Python-native agent runner. For Hermes Agent, use the dedicated Hermes Agent page. Its primary integration is the agentskills.io skill install path, not the patterns below.

Run under the proxy

authsome login github
authsome login openai
authsome run -- python -m <your_runner>

The runner sees placeholder env vars. Outbound HTTPS to matched providers is authenticated at the proxy layer. This is the recommended path.

Export when the proxy is not enough

If you need explicit env vars in the runner process:

eval "$(authsome export github --format env)"
eval "$(authsome export openai --format env)"
python -m <your_runner>

See Python library for the daemon HTTP API.

When to use which

Use the proxy	Use export
You want a single layer that catches every outbound call.	You need fine-grained control over which connection to use per agent.
You’re using third-party clients you can’t easily modify.	The agent is yours and reads env vars directly.
You need transparent injection for shell tools the agent spawns.	You don’t spawn shell tools.

What’s next

Run agents with the proxy

The proxy injection model.

Python library

the CLI export path or the daemon HTTP API.

Cowork Hermes Agent

Agent frameworks

OAuth providers

API-key providers

NanoClaw, OpenClaw

Run under the proxy

Export when the proxy is not enough

When to use which

What’s next

Run agents with the proxy

Python library

​Run under the proxy

​Export when the proxy is not enough

​When to use which

​What’s next

Run agents with the proxy

Python library

Run under the proxy

Export when the proxy is not enough

When to use which

What’s next