// docs / integrations

OpenClaw

Local OpenClaw gateway. Your self-hosted agent speaks through a Primeta avatar.

Route the replies from your locally-hosted OpenClaw agent — Kimi, Llama, GPT, Claude, whatever you've configured — through a Primeta avatar.

When to use this

  • You run a local OpenClaw gateway with your own model
  • You want the agent's text replies spoken through a 3D avatar with voice and emotion
  • You need persona switching to drive the agent's personality mid-conversation

Setup

Generate your Primeta API token at Settings → Token Authentication.

Then install the channel plugin from ClawHub, point it at your token, and restart the gateway:

openclaw plugins install clawhub:openclaw-plugin-primeta
openclaw primeta init --token YOUR_TOKEN --name my-project
openclaw gateway restart

A green OpenClaw card appears on your Primeta dashboard when the gateway connects.

How it works

OpenClaw is a channel connection (not an MCP session). The plugin opens a persistent WebSocket from your local OpenClaw gateway to Primeta — Primeta runs remotely, so the plugin has to initiate. When Primeta forwards a user message, the plugin dispatches one turn through OpenClaw's reply pipeline, buffers the output, and sends a single reply frame back. Personas are injected as a cacheable system-prompt prefix, so the agent stays in character across turns without re-shipping the full personality every message.

Because it's a channel, you can send messages into it from the Primeta chat composer — MCP sessions don't accept typed messages from the UI today.

See sessions, connections, conversations for how sessions work once connected, including live persona switching.

← Back to all docs