OpenCode Bridge for Copilot Chat

Use OpenCode Console and OpenCode Go models directly from the GitHub Copilot Chat model picker in Visual Studio Code.

This native VS Code LanguageModelChatProvider registers Console and Go as separate provider groups and streams their responses into Copilot Chat without a local proxy. VS Code 1.125 exposes the provider-entry configuration through the built-in Language Models UI; the packaged extension does not enable proposed APIs.

Highlights

• Separate OpenCode Console and Go model groups
• Service-account API keys stay in each VS Code provider entry; the extension manages only device-code accounts in Secret Storage
• Named accounts that serve both Console and Go from one device session
• Credential-scoped live discovery, with six-hour persisted models.dev enrichment for Console and Go
• Streaming text, reasoning, image inputs, and agent-mode tool calls
• Thinking stays before the answer or tool calls, with parallel tool arguments and call IDs preserved across interleaved stream events.
• Model-specific Thinking Effort and Qwen thinking-budget controls
• Bounded gateway retries and context-aware token limits
• Provider-entry-scoped inference-token tracking and secret-safe diagnostics

Quick start

1. Install OpenCode Bridge for Copilot Chat from the Visual Studio Marketplace. You need VS Code 1.125 or newer and GitHub Copilot Chat.
2. Alternatively, run code --install-extension grikomsn.opencode-bridge-for-copilot-chat. For manual installation, download the .vsix from the latest GitHub release and run code --install-extension ./opencode-bridge-for-copilot-chat-<version>.vsix --force.
3. For Console or Go, open Manage Language Models, choose Add Models, select the provider, name the entry, set a unique lowercase entryId, and paste its service-account API key. Both Console and Go entries can instead reference an OpenCode account by profile ID: Console may use a device-code sign-in, and Go accepts either a device-code profile or an account-scoped Go key.
4. For device-code sign-in, run OpenCode: Add Console Account, choose a profile ID, then add a Console (or Go) entry in Manage Language Models with the same profile ID and no API key.
5. Optional: give key-only entries a stable Entry Label so their model IDs and usage scope survive key rotation.
6. Repeat any flow for another account or key — for example Console on a personal account and Go on a work subscription — then enable the models you want in Copilot Chat. Review entries and accounts in the Manage Connection menu keeps the two sides in step.

Marketplace installations receive updates through VS Code. Manual VSIX installations can be updated by installing the newest release asset.

Moving from the previous extension

Version 1.0.0 starts a new listing with ID grikomsn.opencode-bridge-for-copilot-chat. Disable or uninstall grikomsn.opencode-bridge-copilot-chat before installing the new extension to avoid duplicate providers and commands. Sign in again or re-enter your API keys in the new extension; its Secret Storage and local usage state are separate. Existing opencode.* workspace settings retain their names.

Composer controls override workspace defaults; ordered effort controls default to High, binary controls default On, and Qwen defaults Auto. Console and Go use each entry's authenticated live catalog, while a Console device-code profile shows only models enabled for that profile's selected organization. Click the OpenCode status-bar item to inspect tokens for the most recently used entry.

Documentation

• Setup, commands, settings, and troubleshooting
• Models and pricing
• Credential handling and security
• Development and releases

Related projects

• Codex Bridge for Copilot Chat
• Grok for GitHub Copilot Chat
• Ollama Cloud for GitHub Copilot Chat
• Poolside for GitHub Copilot Chat

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