Skip to content
inferamdocs
Website ↗

Editors

VS Code

Use Inferam in Copilot Chat's model picker and run the Claude Code extension on Inferam.

  • VS Code 1.144 or later.
  • Signed in to GitHub Copilot in VS Code. Click Accounts (bottom left) → Sign in with GitHub to use GitHub Copilot. The free Copilot plan is enough; you don’t need a paid subscription. Copilot Chat only offers models you bring once you’re signed in: without it, there’s no Chat: Manage Language Models command and chat answers “Language model unavailable”.
  • On a Copilot Business or Enterprise account, your organization must allow “Bring Your Own Language Model Key in VS Code” (a Copilot policy). The free plan and Copilot Pro allow it.
  • For Claude Code inside VS Code, the Claude Code extension. It runs on Inferam; you don’t need an Anthropic account.
  • The Inferam CLI. Setup signs you in if this computer isn’t signed in yet.
Terminal window
inferam setup vscode

Open the Chat view: new chats start on DeepSeek-V4.1-Flash-1017 (Inferam). It works in Ask, Edit and Agent modes, and VS Code picks it up while it’s running, so there’s nothing to restart. If you use the Claude Code extension, its next session runs on Inferam too.

Two files in your VS Code user folder: ~/Library/Application Support/Code/User on macOS, ~/.config/Code/User on Linux, %APPDATA%\Code\User on Windows. Everything else in them, comments included, is left as it is.

chatLanguageModels.json: Copilot Chat’s model

Section titled “chatLanguageModels.json: Copilot Chat’s model”

An “Inferam” entry in Copilot Chat’s Custom Endpoint provider. It’s what Chat: Manage Language Models lists under Inferam.

{
"vendor": "customendpoint",
"name": "Inferam",
"apiType": "messages",
"models": [
{
"id": "DeepSeek-V4.1-Flash-1017",
"name": "DeepSeek-V4.1-Flash-1017 (Inferam)",
"url": "https://api.inferam.ai/v1/messages",
"toolCalling": true,
"vision": true,
"thinking": true,
"contextWindow": 262144,
"maxInputTokens": 229376,
"maxOutputTokens": 32768,
"requestHeaders": { "x-api-key": "ifr_…" }
}
]
}

The key is VS Code’s own key from inferam.ai/keys, sent with each request. The file is made readable only by you.

{
"chat.defaultModel": "DeepSeek-V4.1-Flash-1017",
"claudeCode.claudeProcessWrapper": "/Users/you/.local/bin/inferam",
"claudeCode.disableLoginPrompt": true
}

chat.defaultModel makes new chats start on Inferam’s model; --undo puts back the default you had. The extension starts Claude through inferam, exactly as inferam claude does in a terminal, so it runs on Inferam with this computer’s sign-in. Claude Code in your terminal is unchanged.

“Language model unavailable”, or no Manage Language Models command

Section titled ““Language model unavailable”, or no Manage Language Models command”

You’re not signed in to GitHub Copilot in VS Code. Sign in (Accounts, bottom left; the free plan is enough), and the Inferam model shows up in the Chat view.

Copilot Chat removes custom models when you’re signed in to GitHub with an account that has no Copilot, or when your organization’s Copilot Business or Enterprise policy turns off “Bring Your Own Language Model Key in VS Code”. To confirm, open View → Output and pick GitHub Copilot Chat. You’ll see:

BYOK: unregistered providers because the Copilot token needed to check the BYOK policy is unavailable.

or “…because client BYOK is not allowed by policy.” Sign in with an account that has Copilot (the free plan is enough), or ask your organization’s admin to allow the policy, then run Developer: Reload Window.

Same cause as above: Copilot Chat has unregistered its custom models. Check Output → GitHub Copilot Chat as described there.

Run inferam setup vscode again. It replaces VS Code’s key with a new one, which also fixes a key revoked on inferam.ai/keys. The error reference explains the API’s errors.

The Claude Code extension asks you to sign in

Section titled “The Claude Code extension asks you to sign in”

Check that claudeCode.claudeProcessWrapper in settings.json points at inferam (which inferam) and that inferam auth status shows you signed in. Run inferam setup vscode again after moving or reinstalling inferam.

  • Inline suggestions, codebase indexing and semantic search stay on GitHub’s service. Copilot doesn’t let a custom model serve them.
  • In Agent mode, each request carries VS Code’s tool definitions and instructions. Long agent sessions compact as usual within the 262,144-token window.
Terminal window
inferam setup vscode --undo

Removes the Inferam entry and the Claude Code settings, puts back your default chat model, and revokes VS Code’s key.