OpenClaw

This guide is best suited for Linux, macOS, or remote-server OpenClaw setups.

1. Install OpenClaw

On Linux or macOS, install OpenClaw with:

curl -fsSL https://openclaw.ai/install.sh | bash

Follow the installer prompts and wait for the first-run onboarding to appear.

2. Finish the first-run onboarding

Read the security prompt, confirm it, and choose QuickStart.

If OpenClaw asks you to pick a built-in provider during onboarding, choose Skip for now.

3. Add FHRouter connection details

Use the adapter that matches the model family:

Model family	Adapter	Base URL	Example model
Claude	Anthropic	https://api.fhrouter.com	claude-sonnet-4-6
GPT/Codex	OpenAI-compatible	https://api.fhrouter.com/v1	gpt-5.4-mini
Gemini	Gemini	https://api.fhrouter.com/v1beta	gemini-3.1-pro-preview

Use your FHRouter API key as the credential.

If OpenClaw asks for a model name, copy the exact id from Endpoints and Models.

4. Restart the gateway

After changing provider settings, restart the OpenClaw gateway:

openclaw gateway restart

5. Test in TUI

Run:

openclaw tui

Send one short prompt first. For example:

Reply with one short sentence confirming the connection works.

6. Optional browser access

If you expose the OpenClaw dashboard on a remote server, put it behind your own reverse proxy and TLS. Do not leave a raw management endpoint public without protection.

Common issues

• If OpenClaw cannot find the model, copy the exact model id again from FHRouter.
• If the adapter expects OpenAI-compatible traffic, include /v1.
• If the adapter expects Claude or Anthropic traffic, use the root host without /v1.
• If the adapter expects Gemini-native traffic, use /v1beta.
• After any provider or model change, restart the gateway before testing again.