Deetoo brings Nasdaq-provided AI models into GitHub Copilot Chat in VS Code.
It is built for teams that want to stay inside the standard Copilot Chat experience while using an approved provider in the editor.
Quick start
Install GitHub Copilot Chat in VS Code.
Install Deetoo from the Marketplace.
Create or update an OAuth2 client in the matching GenAI token page for your environment. In dev, include the Gateway Access scope. In prod, make sure the client has been provisioned for Bedrock gateway access.
Run Deetoo: Set Gateway Credentials from the Command Palette (Cmd/Ctrl+Shift+P).
Enter that client ID and client secret when prompted.
Open Copilot Chat and choose a Deetoo model from the model picker.
Why Deetoo
Use an approved provider without leaving VS Code
Keep the familiar Copilot Chat workflow, UI, and history
Give teams a simple way to access Nasdaq-supported AI models in the editor
What you can do
Select Deetoo models from the Copilot Chat model picker
Chat in the same VS Code interface your team already uses
Manage credentials directly from the Command Palette
Commands
Command
Purpose
Deetoo: Set Gateway Credentials
Connect Deetoo with your access credentials.
Deetoo: Clear Gateway Credentials
Remove stored credentials.
Deetoo: Open Settings
Jump to Deetoo settings UI.
Deetoo: Show Logs
Open the Deetoo output channel.
Notes
Deetoo works alongside GitHub Copilot Chat and appears in the standard model picker.
If your team has onboarding or access requirements, follow your internal setup guidance.
PROD offers two model sources: bedrock (default, 5 Anthropic Claude models, SigV4, with quota display) and openai-proxy (30 models including Amazon Nova, Anthropic Claude, Google Gemini, and open-weight models — NVIDIA Nemotron, OpenAI GPT-OSS, Google Gemma, xAI Grok). Switch with the Deetoo: Use Nasdaq OpenAI Proxy Models (all models, incl. OSS) / Deetoo: Use Bedrock Gateway Models commands.
Tool calling works reliably on Claude, Nova, and Gemini models through both providers. openai/gpt-oss-* and nvidia/nemotron-* models can emit malformed tool calls (unknown_tool name, empty arguments) through the openai-proxy path — a known upstream limitation, not a Deetoo bug. Deetoo now skips malformed tool calls, keeps any streamed text, retries once if the turn was empty because of a malformed tool call, and then shows guidance instead of failing the request. Set deetoo.toolMode to off if you want to use OSS models in chat-only mode.
Troubleshooting
If Deetoo does not appear after install, reload VS Code with Developer: Reload Window.
If no models appear or requests fail with gateway-access, make sure the OAuth2 client used by Deetoo has Bedrock gateway access. In dev that means selecting the Gateway Access scope; in prod that access appears to be provisioned on the client.
