Poolside for GitHub Copilot Chat

Use hosted Poolside coding models directly from the GitHub Copilot Chat model picker in Visual Studio Code.

This extension is a native VS Code LanguageModelChatProvider. It validates a user-supplied Poolside Platform API key, discovers the hosted models available to that key, and streams responses from inference.poolside.ai into Copilot Chat without a local proxy.

Highlights

• Direct, first-class Poolside Platform integration
• API keys stored in VS Code Secret Storage
• Live hosted-model discovery with sensible Laguna fallbacks
• Streaming text and model reasoning
• Configurable reasoning effort in the Copilot model picker (none, minimal, low, medium, high, or xhigh)
• Agent mode function-tool calls
• Native VS Code context-window accounting from Poolside usage data
• No proxy, bundled server, or third-party relay

Quick start

1. Install Poolside for GitHub Copilot Chat. You need VS Code 1.125 or newer and GitHub Copilot Chat.
2. Create a developer API key in Poolside Platform.
3. Run Poolside: Configure API Key from the Command Palette. The extension validates the key before saving it.
4. Open Copilot Chat, select Manage Models, enable Poolside, then choose an available Laguna model.

Use Poolside: Manage Connection to test inference, refresh hosted models, replace or remove the key, inspect logs, or create a diagnostic snapshot.

Choose a reasoning level from the model configuration control in Copilot Chat. The selection applies to that request and overrides the poolsideCopilot.reasoningEffort workspace default.

Poolside's hosted Laguna models are currently text-only. Image attachments are disabled; prompts, tool definitions, tool results, and conversation context selected by Copilot Chat are sent directly to Poolside for inference.

Documentation

• Setup, settings, and troubleshooting
• API key and security model
• Development and releases

Unofficial project; not affiliated with Poolside, GitHub, or Microsoft. Poolside account limits and charges still apply. Licensed under MIT.