Your workflow. Your model.

The current Shard build includes several ways to connect an AI model. Launch availability is still to be confirmed.

Connect an account.

The current build includes two connected-account providers:

  • Codex uses your ChatGPT account. Settings gives you a code to enter in your browser, then shows the account in use.
  • Claude Code uses your Claude account. You run claude setup-token on your own computer and paste the token into Settings.

These connections are still in development. Their availability depends on the integration, the connected account and the provider’s terms, and which plans qualify isn’t confirmed. Shard’s launch setup will be documented when it is ready.

Bring an API key.

The development catalog includes OpenAI, Anthropic, Gemini, xAI, Z.AI, OpenRouter, Amazon Bedrock, Moonshot, Meta and Groq. Each provider has its own model access and usage terms.

Keys are saved in Studio’s plugin settings. Relevant conversation content and the key are sent through Shard infrastructure to make the provider request.

Shard-hosted models.

Choose the Shard provider and sign in to use hosted models with Shards. No API key is needed. The current catalog lists Shard: GPT-6.1 Sol. The launch catalog and usage limits have not been announced.

Settings that fit the conversation.

Each chat keeps its provider and model selection. Supported models can offer reasoning settings, context controls and prompt caching. These options depend on the chosen provider and model.

Read the provider preview and understand the data flow.