AI Agent Hub
Back to plugins
🤖

dsh-xai-catalog

Model Inference Updated 2026.08.22

Run the following command in DeepSeek Harness:

dsh plugin install OpenSaozi/dsh-xai-catalog

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install OpenSaozi/dsh-xai-catalog in your DeepSeek Harness profile to add this plugin; source at https://github.com/OpenSaozi/dsh-xai-catalog .

About this plugin

Using xAI Grok inside a DeepSeek Harness workspace has meant hand-wiring endpoints, juggling two protocol families, and babysitting OAuth token expiry. dsh-xai-catalog collapses that friction into a single plugin: on mount it queries the xAI /models endpoint so every Grok model available to your account appears in the model picker in real time; grok-4.3 and grok-build-0.1 are routed through Chat Completions while grok-4.5 and grok-4.6 go through Responses, all handled by a dedicated xai adapter registered by the plugin. For credentials, it reads the JSON already stored at XAI_OAUTH_CREDENTIAL, delegates refresh and rotation to pi-ai's built-in xAI OAuth handler, and writes the new token back through the Harness credential service — the raw JSON never touches Cordis config or logs, and only the handler-derived access token reaches the adapter as a per-request auth override. The integration injects no extra prompts, headers, or transport changes, keeping token overhead at zero. It is built for teams and individual developers who hold an xAI API key or subscription and want Grok 4.6's coding and reasoning capabilities available alongside their existing DeepSeek Harness profile, with no local build steps required.

Use Cases

  • Invoke Grok 4.6 for advanced coding and reasoning tasks inside a DeepSeek Harness workspace
  • Switch between grok-4.3, grok-4.5, and grok-4.6 from the model picker without editing endpoint config
  • Auto-refresh expired OAuth tokens so long-running inference sessions never break on auth failure

Best For

  • Developers who hold an xAI API key or subscription and want Grok's coding capabilities alongside their existing stack
  • Teams running multi-model orchestration in DeepSeek Harness that need a unified external LLM integration
  • Platform engineers who want a zero-build, drop-in path to add a new model provider to their pipeline