Skip to content
| Marketplace
Sign in
Visual Studio Code>Machine Learning>GrawlixNew to Visual Studio Code? Get it now.
Grawlix

Grawlix

Preview

grawlix

|
3 installs
| (2) | Free
Route GitHub Copilot Chat through the Grawlix proxy with one org-issued key. Anthropic, OpenAI, and Gemini models.
Installation
Launch VS Code Quick Open (Ctrl+P), paste the following command, and press enter.
Copied to clipboard
More Info

Grawlix

Route GitHub Copilot Chat through the Grawlix proxy with a single organization-issued API key — Anthropic, OpenAI, and Gemini models, all behind one key.

The extension registers Grawlix as a Copilot language-model provider, so you pick a (Grawlix) model in the Copilot Chat model picker and requests route through your organization's proxy. You authenticate with one Grawlix API key; the underlying Anthropic / OpenAI / Gemini keys stay in the proxy and never reach your machine.

Setup

  1. Install the extension.
  2. Run Grawlix: Manage from the Command Palette (Ctrl/Cmd + Shift + P) and enter your organization-issued Grawlix API key. It's stored in VS Code's OS-backed Secret Storage.
  3. Open Copilot Chat and choose a (Grawlix) model from the model picker.

Copilot Ask and Edit work with Grawlix models. Inline code completions (grey ghost text) are unaffected and continue to run on GitHub's infrastructure.

Models

Every model supports tool calling and image input.

Model Provider Context window Max output
Claude Fable 5 Anthropic 1M 128K
Claude Opus 5 Anthropic 1M 128K
Claude Opus 4.8 Anthropic 1M 128K
Claude Sonnet 5 Anthropic 1M 128K
Claude Sonnet 4.6 Anthropic 1M 128K
GPT-5.6 Sol OpenAI 1.05M 128K
GPT-5.6 Terra OpenAI 1.05M 128K
GPT-5.6 Luna OpenAI 1.05M 128K
GPT-5.5 Pro OpenAI 1.05M 128K
GPT-5.5 OpenAI 1.05M 128K
GPT-5.4 OpenAI 1.05M 128K
GPT-5.4 mini OpenAI 400K 128K
GPT-5.3 Codex OpenAI 400K 128K
Gemini 3.1 Pro (Preview) Gemini 1M 64K
Gemini 3.6 Flash Gemini 1M 64K
Gemini 3.5 Flash Gemini 1M 64K
Gemini 3.5 Flash-Lite Gemini 1M 64K

GPT-5.5 Pro does not stream: its answer appears all at once when the request completes, rather than token by token. It also ignores the low setting of grawlix.reasoningEffort, treating it as medium (the lowest rung it accepts).

Models appear in the picker with a (Grawlix) suffix. The exact list available to you depends on your organization's proxy configuration.

  • Contact us
  • Jobs
  • Privacy
  • Manage cookies
  • Terms of use
  • Trademarks
© 2026 Microsoft