Provides speech-to-text transcription using Deepgram's free API.
Provides image generation using Flux's free tier via API key.
Provides chat and vision analysis using Gemini models (3.5 Pro, 3.5 Flash) via OAuth token authentication.
Provides video generation using Kling models (1.6) via browser session cookie authentication.
Provides chat completion using ChatGPT models (GPT-4o, GPT-4.5) via browser session cookie authentication.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@finalyy generate an image of a serene mountain lake"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
No API keys. No subscriptions. Just point your IDE and go.
Finalyy converts free browser-based AI trials (ChatGPT, Claude, Gemini, DeepSeek, Flux, Kling, and more) into MCP tools and OpenAI-compatible endpoints. One CLI command, zero cost, production-grade quality.
npx finalyy init # One-time setup
npx finalyy start # Launch MCP server on localhost:1337Then connect any MCP client — Cursor, Claude Desktop, Codex, Continue.dev, Cline, OpenClaw — and use frontier models for free.
Every week, new AI providers launch free browser trials of frontier models. Developers can use them in a browser but cannot integrate them into their tools, IDEs, or agents because:
Providers charge for API keys ($50–$500/mo)
Browser sessions don't expose OpenAI-compatible endpoints
No standard way to convert web trials → MCP tools
Each provider has a different authentication mechanism
Finalyy solves this — one unified MCP server, community-driven provider registry, auto-fallback, zero API keys.
Related MCP server: Browser Controller
Node.js 18+
A Chromium-based browser (Chrome, Edge, Brave)
An active free account on any supported AI provider
npx finalyy initThis will:
Install the Finalyy CLI
Create ~/.finalyy/config.json
Open a browser page to log into your AI providers
Capture and encrypt session cookies/tokens
Verify connectivity
npx finalyy startMCP server starts at http://localhost:1337. Connect any MCP client.
Cursor:
// .cursor/mcp.json
{
"mcpServers": {
"finalyy": {
"command": "npx",
"args": ["finalyy", "start"]
}
}
}Claude Desktop:
// claude_desktop_config.json
{
"mcpServers": {
"finalyy": {
"command": "npx",
"args": ["finalyy", "start"]
}
}
}Claude Code:
claude mcp add npx finalyy startContinue.dev / Cline / Codex:
// .vscode/mcp.json or config.json
{ "mcpServers": { "finalyy": { "command": "npx", "args": ["finalyy", "start"] }}}curl http://localhost:1337/v1/models{
"models": [
{ "id": "chatgpt/gpt-4o", "provider": "chatgpt", "type": "chat", "status": "online" },
{ "id": "claude/sonnet-4", "provider": "claude", "type": "chat", "status": "online" },
{ "id": "gemini/pro-3.5", "provider": "gemini", "type": "chat", "status": "online" },
{ "id": "flux/pro", "provider": "flux", "type": "image", "status": "online" },
{ "id": "kling/1.6", "provider": "kling", "type": "video", "status": "online" }
]
}Tool | Description | Provider Examples | Model Types |
| Chat completion | ChatGPT, Claude, Gemini, DeepSeek | text, code, reasoning |
| Image generation | Flux, DALL-E free, Stable Diffusion | text→image |
| Video generation | Kling, Runway, Pika | text→video |
| Speech-to-text | Whisper, Deepgram free | audio→text |
| Vision analysis | Gemini Vision, Claude Vision | image+text→text |
Each tool maps to a specific provider's free tier through Finalyy's session adapter layer.
Provider | Models | Type | Session Method | Status |
ChatGPT | GPT-4o, GPT-4.5 | chat | Cookie auth | ✅ Active |
Claude | Sonnet 4, Opus 4.6 | chat, vision | Cookie auth | ✅ Active |
Gemini | 3.5 Pro, 3.5 Flash | chat, vision | OAuth token | ✅ Active |
DeepSeek | V3.2, R1 | chat, code | Cookie auth | ✅ Active |
Flux | Pro 1.2, Dev | image | API key (free) | ✅ Active |
Kling | 1.6 | video | Cookie auth | ✅ Active |
100+ providers in community registry. See Registry docs for adding new providers.
┌──────────────────────────────────────────────────┐
│ YOUR IDE / AGENT │
│ (Cursor · Claude Desktop · Codex · Continue) │
└──────────────────────┬───────────────────────────┘
│ MCP Protocol (localhost:1337)
┌──────────────────────▼───────────────────────────┐
│ FINALYY │
│ │
│ ┌──────────┐ ┌──────────┐ ┌────────────────┐ │
│ │ CLI │ │ MCP │ │ Dashboard │ │
│ │ (Init/Add │ │ Server │ │ (Web UI) │ │
│ │ /Start) │ │ (Tools + │ │ Provider status│ │
│ │ │ │ Resources│ │ Usage + logs │ │
│ └──────────┘ └────┬─────┘ └────────────────┘ │
│ │ │
│ ┌──────────────────▼──────────────────────────┐ │
│ │ Session Manager │ │
│ │ ┌────────┐ ┌────────┐ ┌────────┐ ┌──────┐ │ │
│ │ │ChatGPT │ │ Claude │ │ Gemini │ │ ... │ │ │
│ │ │Session │ │Session │ │Session │ │ │ │ │
│ │ └────────┘ └────────┘ └────────┘ └──────┘ │ │
│ │ Cookie Rotator · Health Checker · Captcha │ │
│ └──────────────────┬──────────────────────────┘ │
└─────────────────────┬─────────────────────────────┘
│ HTTP/WebSocket (per provider)
┌─────────────────────▼─────────────────────────────┐
│ AI PROVIDERS (Free Tiers) │
│ chatgpt.com · claude.ai · gemini.google.com │
│ deepseek.com · flux.ai · kling.kuaishou.com │
└───────────────────────────────────────────────────┘Metric | Finalyy | gpt4free | AIClient2API | OpenProxy |
Providers | 20+ built-in · 100+ community | 60+ | 12+ | 40+ (API keys only) |
MCP Support | ✅ Native | ❌ | ❌ | ❌ |
Multi-model types | chat, image, video, audio | chat only | chat only | chat only |
Browser sessions | ✅ Cookie/OAuth capture | ❌ (scraping) | ❌ (CLI apps) | ❌ (API keys) |
Session health | ✅ Auto-rotate, captcha detect | ❌ | ❌ | ❌ |
Region routing | ✅ Proxy per provider | ❌ | ❌ | ❌ |
Community registry | ✅ Auto-updating PR system | ❌ | ❌ | ❌ |
Encryption at rest | ✅ AES-256-GCM | ❌ | ❌ | ❌ |
Startup time | < 500ms | — | — | — |
Memory usage | ~80 MB idle | ~200 MB | ~150 MB | ~30 MB |
Response latency | +150ms (proxy overhead) | +300-800ms | +200-400ms | +50ms |
Finalyy = Final + Why.
"The final answer to 'why do I still need to pay for AI APIs?'"
MIT — free for everyone, forever.
Always free. We don't charge. We don't have a paid tier. We never will.
Always open. Source code, provider registry, community — all open.
Always current. Providers change. We adapt. Community helps.
Never scam. No hidden fees. No crypto. No data collection. Just MCP.
This server cannot be deployed
Connect MCP clients to 2,000+ AI models without managing provider API keys.
Zero-setup MCP gateway securely connecting AI to your tools with authentication and workflows
Free OpenAI-compatible inference with signed provenance receipts and 3 focused MCP tools.
One MCP endpoint for Claude, GPT & Gemini: 100+ tools + no-code connectors + agent workers.
