Show HN: Turn your Google accounts into a free, load-balanced LLM API gateway

OpenGem, an open-source tool, allows users to turn one or more Google accounts into a local, load-balanced LLM API gateway that supports native Gemini, OpenAI, and Anthropic endpoints. The project, designed for personal, educational, and research use, provides a Next.js admin console for managing API keys, monitoring usage, and rotating accounts with automatic quota handling. By running the gateway locally, users can create their own API keys and route requests through multiple Google accounts, effectively creating a free, load-balanced alternative to paid API services.

OpenGem turns one or more Google accounts into a local, load-balanced Gemini gateway. It exposes the native Gemini API shape and also accepts OpenAI Chat Completions and Anthropic Messages requests, so most SDKs can point at your OpenGem server with only a base URL change. It is designed for personal, educational and research usage. You run the gateway, connect your own Google accounts, create your own API keys, and monitor usage from the admin console. Next.js admin console with a rebuilt setup wizard, dashboard, logs, API keys, docs, settings and chat playground. Native Gemini endpoint : POST /v1beta/models/{model}:generateContent . OpenAI-compatible endpoint : POST /v1/chat/completions and GET /v1/models . Anthropic-compatible endpoint : POST /v1/messages . Multi-account rotation with cooldowns, retries, quota handling and automatic account reactivation. Local SQLite backend using Node.js node:sqlite , plus optional Firebase Firestore. Secure API key store with hashed keys, JWT admin sessions, rate limiting and Helmet headers. Streaming support across Gemini, OpenAI-compatible and Anthropic-compatible surfaces. Function calling, system prompts, tool payloads, image payloads and request logging . No legacy : the UI is now built through Next static export and served from public/ frontend out/ . - Node.js 22.5.0 or newer - npm - At least one Google account - Optional: Firebase project, only if you choose Firestore instead of local SQLite git clone https://github.com/arifozgun/OpenGem.git cd OpenGem npm install npm run build npm start Open http://localhost:3050 . On a fresh install, OpenGem redirects to /setup where you choose the database backend and create the admin account. After setup, sign in, connect Google accounts, then create an API key from the dashboard. For development: npm run dev The default server bind is 127.0.0.1:3050 . Use HOST=0.0.0.0 only when you intentionally expose the process behind your own network controls. All endpoints accept the same OpenGem API keys. Authorization: Bearer sk-your-api-key x-goog-api-key: sk-your-api-key x-api-key: sk-your-api-key ?key=sk-your-api-key curl -X POST "http://localhost:3050/v1beta/models/gemini-3.1-pro-preview:generateContent?key=sk-your-api-key" \ -H "Content-Type: application/json" \ -d '{"contents": {"parts": {"text":"Hello from OpenGem"} } }' js import { GoogleGenAI } from "@google/genai"; const ai = new GoogleGenAI { apiKey: "sk-your-api-key", baseUrl: "http://localhost:3050", } ; const response = await ai.models.generateContent { model: "gemini-3.1-pro-preview", contents: "Explain OpenGem in one paragraph.", } ; console.log response.text ; python import OpenAI from "openai"; const client = new OpenAI { apiKey: "sk-your-api-key", baseURL: "http://localhost:3050/v1", } ; const response = await client.chat.completions.create { model: "gpt-4o", messages: { role: "user", content: "Hello " } , } ; console.log response.choices 0 .message.content ; python import Anthropic from "@anthropic-ai/sdk"; const client = new Anthropic { apiKey: "sk-your-api-key", baseURL: "http://localhost:3050", } ; const response = await client.messages.create { model: "claude-3-5-sonnet-latest", max tokens: 1024, messages: { role: "user", content: "Write a short haiku." } , } ; console.log response.content ; The dashboard includes: Overview : request totals, success rate, active accounts and token usage. Accounts : connect Google accounts, reactivate cooled-down accounts and remove accounts. API Keys : create, copy-once and revoke gateway keys. Request Logs : inspect prompt, response, model, fallback and token metadata. Documentation : endpoint quick reference and an API playground. Chat : authenticated Gemini chat with streamed responses and thinking sections. Settings : database backend switching, credential rotation and privacy mode. OpenGem stores runtime configuration in config.json after setup. Local data lives in data/db.sqlite when SQLite is selected. Both are ignored by git. Useful environment variables: PORT=3050 HOST=127.0.0.1 CORS ORIGIN=https://your-domain.example When using Firebase, paste the Web app config in the setup wizard or in Settings when switching backends. The backend is still Express + TypeScript, while the UI is a static Next.js export. npm run build:server compiles src/ to dist/ npm run build:web exports app/ to out/ npm run build runs both npm start serves API routes and the built Next UI from the same port. OpenGem/ ├── app/ Next.js admin console and setup wizard │ ├── assets/ UI images/icons imported by Next │ ├── setup/ Setup wizard route │ └── opengem-console.jsx Dashboard application ├── components/ui/ shadcn-style UI primitives ├── lib/ Frontend utilities ├── src/ │ ├── controllers/ Gemini, OpenAI and Anthropic handlers │ ├── middleware/ Admin auth middleware │ └── services/ Config, database, OAuth, streaming, retry logic ├── dist/ Compiled backend output ├── out/ Next static export output ├── data/ Local SQLite data, ignored by git ├── config.json Runtime config, ignored by git ├── app.js Production entry for Passenger/cPanel └── package.json - Keep config.json , .env and data/ private. - Place production deployments behind nginx, Cloudflare or another trusted reverse proxy. - Keep the Node process bound to loopback unless you know why it must be exposed. - Regenerate API keys after switching database backends; raw key material is intentionally not recoverable. MIT. See LICENSE /arifozgun/OpenGem/blob/main/LICENSE .