OpenRouter Server — VS Code Extension

Route Claude Code CLI through alternative LLM providers (OpenRouter, NVIDIA NIM, or LM Studio) from your VS Code terminal.

What It Does

This VS Code extension bundles a local FastAPI proxy server that intercepts Anthropic API calls from the Claude Code CLI and forwards them to your choice of provider:

OpenRouter — access 200+ models from Anthropic, Meta, Mistral, etc. at usage-based pricing
NVIDIA NIM — run models locally or via NVIDIA's microservices platform
LM Studio — route to a local LM Studio instance on localhost:1234

The extension provides a status bar toggle (default: off) that starts/stops the proxy and automatically injects the required environment variables into every VS Code terminal so Claude Code CLI works without manual configuration.

How It Works

Toggle ON (status bar item):
- Spawns a local uvicorn/FastAPI server in server/
- Injects env vars (ANTHROPIC_BASE_URL, OPENROUTER_API_KEY, model mappings) into all new terminals
- Claude Code CLI automatically routes through the proxy
Make requests via Claude Code CLI — the proxy maps Claude model names (claude-opus-4, claude-sonnet-4-5, etc.) to your configured provider/model strings
Toggle OFF:
- Stops the server
- Clears injected env vars

The status bar shows the proxy state and, after your first request, the active model and tier being used.

Prerequisites

Before you begin, ensure you have:

VS Code 1.90 or later
Node.js 16+ and npm (to build the extension)
Python 3.14+ (for the proxy server)
uv package manager — install from https://github.com/astral-sh/uv
At least one API key for your chosen provider:
- OpenRouter API key (free to sign up)
- NVIDIA NIM API key (free tier available)
- A running LM Studio instance on localhost:1234 (free, open-source)

Installation

1. Build & Install the VS Code Extension

# Clone the repository
git clone <your-repo-url>
cd openrouter-vscode-extension

# Install Node dependencies
npm install

# Build the extension
npm run compile

# Package as .vsix
npm run package

This creates openrouter-vscode-*.vsix in the root directory.

In VS Code:

Press Ctrl+Shift+P → "Extensions: Install from VSIX..."
Select the .vsix file

(Alternatively, drag the .vsix into VS Code's Extensions panel.)

2. Configure the Proxy Server

# Install Python dependencies
cd server
cp .env.example .env

Edit server/.env with your API keys and model choices:

# Choose your primary provider and add its API key
OPENROUTER_API_KEY=sk-...

# Map Claude model tiers to provider/model strings
MODEL_OPUS=open_router/anthropic/claude-opus-4
MODEL_SONNET=open_router/anthropic/claude-sonnet-4-5
MODEL_HAIKU=open_router/anthropic/claude-haiku-4-5

# (Or use a local LM Studio instance)
# LM_STUDIO_BASE_URL=http://localhost:1234/v1
# MODEL_OPUS=lmstudio/mistral-7b

Then install Python dependencies:

uv sync

3. Configure VS Code Settings (Optional)

Open VS Code Settings → search for "openrouter":

openrouter.apiKey: Your OpenRouter API key (or leave blank to read from server/.env)
openrouter.port: Port for the proxy server (default: 8082)
openrouter.serverPath: Path to the server/ directory (default: bundled in the extension)
openrouter.autoStartServer: Auto-start the server when you toggle ON (default: true)

Usage

Click the "$(circle-slash) OpenRouter" status bar item at the bottom of VS Code to toggle the proxy ON
- First time: the proxy will start (takes a few seconds)
- Status bar shows "$(check) OpenRouter" when running
Open a terminal in VS Code — the required env vars are automatically injected
Use Claude Code CLI normally:
```
claude
```
After your first request, the status bar updates to show the active model and tier:
- "$(check) OpenRouter | Opus"
- "$(check) OpenRouter | Sonnet"
- etc.
Click the status bar item again to toggle OFF (stops the server)

Configuration Reference

VS Code Settings

Setting	Type	Default	Purpose
`openrouter.apiKey`	string	`""`	OpenRouter API key; injected as `OPENROUTER_API_KEY`. Falls back to `server/.env` if empty.
`openrouter.modelOpus`	string	`""`	Provider/model for Claude Opus requests. Format: `provider/model`. Fallback: `openrouter.model`.
`openrouter.modelSonnet`	string	`""`	Provider/model for Claude Sonnet requests. Format: `provider/model`. Fallback: `openrouter.model`.
`openrouter.modelHaiku`	string	`""`	Provider/model for Claude Haiku requests. Format: `provider/model`. Fallback: `openrouter.model`.
`openrouter.model`	string	`""`	Fallback model for all requests. Server default: `nvidia_nim/meta/llama3-70b-instruct`.
`openrouter.port`	number	`8082`	TCP port for the uvicorn proxy server.
`openrouter.serverPath`	string	`""`	Absolute path to a custom `server/` directory. Empty = use bundled.
`openrouter.autoStartServer`	boolean	`true`	Auto-start uvicorn when toggle is enabled.

Server Environment Variables (`server/.env`)

Variable	Example	Purpose
`OPENROUTER_API_KEY`	`sk-...`	Your OpenRouter API key.
`NVIDIA_NIM_API_KEY`	(your key)	Your NVIDIA NIM API key (for `nvidia_nim/` models).
`LM_STUDIO_BASE_URL`	`http://localhost:1234/v1`	Base URL for a local LM Studio instance.
`MODEL_OPUS`	`open_router/anthropic/claude-opus-4`	Provider/model for Claude Opus-tier requests.
`MODEL_SONNET`	`open_router/anthropic/claude-sonnet-4-5`	Provider/model for Claude Sonnet-tier requests.
`MODEL_HAIKU`	`open_router/anthropic/claude-haiku-4-5`	Provider/model for Claude Haiku-tier requests.
`MODEL`	`nvidia_nim/meta/llama3-70b-instruct`	Fallback model for all Claude requests.
`PROVIDER_RATE_LIMIT`	`40`	Max requests per rate window.
`PROVIDER_RATE_WINDOW`	`60`	Rate window in seconds.
`PROVIDER_MAX_CONCURRENCY`	`5`	Max concurrent provider requests.

See server/.env.example for all available configuration options, including bot integration and voice transcription settings.

Model Format

Model strings use the format:

provider/model-name

Valid provider prefixes:

open_router — OpenRouter API
nvidia_nim — NVIDIA NIM
lmstudio — Local LM Studio instance

Examples:

open_router/anthropic/claude-opus-4
open_router/meta-llama/llama-3.1-8b-instruct
open_router/mistral-ai/mistral-7b:free
nvidia_nim/meta/llama3-70b-instruct
nvidia_nim/z-ai/glm4.7
lmstudio/mistral-7b

Optional Features

Voice Transcription

The proxy can transcribe voice notes to text using OpenAI's Whisper model. Choose your transcription backend:

Option 1: NVIDIA NIM Riva (requires NVIDIA account):

cd server
uv sync --extra voice

Then in server/.env:

VOICE_NOTE_ENABLED=true
WHISPER_DEVICE=nvidia_nim
WHISPER_MODEL=openai/whisper-large-v3

Option 2: Local Whisper (CPU or CUDA):

cd server
uv sync --extra voice_local

Then in server/.env:

VOICE_NOTE_ENABLED=true
WHISPER_DEVICE=cuda  # or "cpu"
WHISPER_MODEL=openai/whisper-large-v3

Telegram / Discord Bot Integration

Configure the proxy to receive Claude Code jobs via Telegram or Discord:

In server/.env:

MESSAGING_PLATFORM=discord  # or "telegram"
DISCORD_BOT_TOKEN=...
ALLOWED_DISCORD_CHANNELS=123456,789012

(See server/.env.example for Telegram setup.)

Development

Build Commands

npm run compile      # One-shot TypeScript build
npm run watch        # Watch mode (use during development)
npm run lint         # Run ESLint
npm run test         # Run VS Code extension tests (launches Extension Development Host)
npm run package      # Package as .vsix
npm run deploy       # Publish to VS Code Marketplace

Testing the Extension

Press F5 in VS Code to launch the Extension Development Host — a new VS Code window running your modified extension for manual testing.

Architecture

src/extension.ts — activate/deactivate, status bar toggle, model watcher
src/serverManager.ts — spawns and kills the uvicorn process
src/statusBar.ts — status bar item UI and state
src/envInjector.ts — injects env vars into terminal
src/settingsProvider.ts — reads VS Code settings and server/.env
server/server.py — FastAPI app entry point
server/api/routes.py — Anthropic API route handlers
server/providers/ — OpenRouter, NVIDIA NIM, LM Studio backend implementations

See CLAUDE.md for full developer documentation and data flow diagrams.

Troubleshooting

Port Already in Use

If port 8082 is in use, change it in VS Code settings:

Settings → search "openrouter.port"
Set to an available port (e.g., 8083)

Server Fails to Start

Check the OpenRouter Server output channel in VS Code:

View → Output
Select "OpenRouter Server" from the dropdown
Look for error messages

Common issues:

Missing uv or Python 3.14+: install uv from https://github.com/astral-sh/uv
Missing API keys: edit server/.env
Network timeouts: check your internet connection and provider status

Claude Code CLI Not Using the Proxy

Ensure the extension is ON (status bar shows "$(check) OpenRouter") and you opened the terminal after toggling on — the env vars are only injected to new terminals.

If you opened the terminal before toggling on, close it and open a new one.

License

MIT — see LICENSE for details. Portions of the bundled server are derived from free-claude-code by Alishahryar1, also MIT licensed.

Contributing

Contributions welcome! Please open an issue or pull request.

Built with ❤️ for Claude Code users.

Name		Name	Last commit message	Last commit date
.vscode		.vscode
server		server
src		src
.eslintrc.json		.eslintrc.json
.gitignore		.gitignore
.vscodeignore		.vscodeignore
CLAUDE.md		CLAUDE.md
LICENSE		LICENSE
README.md		README.md
package-lock.json		package-lock.json
package.json		package.json
tsconfig.json		tsconfig.json

Navigation Menu

Repository files navigation

OpenRouter Server — VS Code Extension

What It Does

How It Works

Prerequisites

Installation

1. Build & Install the VS Code Extension

2. Configure the Proxy Server

3. Configure VS Code Settings (Optional)

Usage

Configuration Reference

VS Code Settings

Server Environment Variables (`server/.env`)

Model Format

Optional Features

Voice Transcription

Telegram / Discord Bot Integration

Development

Build Commands

Testing the Extension

Architecture

Troubleshooting

Port Already in Use

Server Fails to Start

Claude Code CLI Not Using the Proxy

License

Contributing

About

Releases

Packages

Contributors

Languages

Latest commit

History

Folders and files

Repository files navigation

OpenRouter Server — VS Code Extension

What It Does

How It Works

Prerequisites

Installation

1. Build & Install the VS Code Extension

2. Configure the Proxy Server

3. Configure VS Code Settings (Optional)

Usage

Configuration Reference

VS Code Settings

Server Environment Variables (server/.env)

Model Format

Optional Features

Voice Transcription

Telegram / Discord Bot Integration

Development

Build Commands

Testing the Extension

Architecture

Troubleshooting

Port Already in Use

Server Fails to Start

Claude Code CLI Not Using the Proxy

License

Contributing

About

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages

Server Environment Variables (`server/.env`)