Open Source · Apache 2.0

One proxy key.
Every LLM model.

Meridian is an open-source LLM API gateway. Add your provider keys once, get a proxy key, and call any model through a unified OpenAI-compatible API. Your real keys never touch your app code.

Before (scattered keys)
// Every file, every service...
const openai = new OpenAI({
  apiKey: "sk-real-key-exposed"
});
const anthropic = new Anthropic({
  apiKey: "sk-ant-another-key"
});
After (one Meridian key)
// One key. Any model.
const res = await fetch(
  "https://your.meridian/api/v1/chat/completions",
  {
    headers: {
      Authorization: "Bearer sk-mrd-..."
    },
    body: JSON.stringify({
      model: "gpt-4.1", // or claude, gemini...
      messages: [{ role: "user", content: "Hi" }]
    })
  }
);

The problem with LLM APIs today

Key Chaos

API keys scattered across .env files, config services, and code. One leak and you're exposed.

Zero Visibility

No idea which team, app, or feature is burning through your LLM budget until the bill arrives.

Vendor Lock-in

Switching from GPT-4 to Claude means rewriting API calls everywhere. No easy way to compare.

Everything you need to manage LLM access

Proxy Keys

Generate revocable proxy keys. Your real API keys stay encrypted and hidden.

AES-256 Encryption

Provider keys encrypted at rest with AES-256-GCM. Decrypted only for the instant of the request.

Unified API

OpenAI-compatible endpoint for all providers. One format, every model.

Analytics & Cost Tracking

Per-request logging with token counts, latency, and cost estimates. Real-time dashboard.

A/B Experiments

Route traffic between models with configurable splits. Compare cost and latency.

100% Open Source

Apache 2.0 licensed. Self-host anywhere. Full control over your LLM infrastructure.

How it works

1

Add your API keys

Connect OpenAI, Anthropic, or Google Gemini. Keys are encrypted instantly.

2

Generate a proxy key

Create a revocable proxy key (sk-mrd-...). Use it in all your apps.

3

Call the latest models

Hit the unified API. Meridian routes, logs, and returns the response.

Supported Providers

More providers coming soon. Contributions welcome.

OpenAI

GPT-4.1, GPT-4.1 Mini, GPT-4o, GPT Image 1

Anthropic

Claude Sonnet 4, Claude Opus 4, Claude 3.7 Sonnet

Google Gemini

Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 2.0 Flash

Built in the open

Meridian is Apache 2.0 licensed. Inspect every line of code. Self-host on your own infrastructure. Contribute features the whole community needs.

LicenseTypeScriptNext.js
View on GitHub