Ask Claude or Cursor which AI models just went live.

ai-model-watch is our free, open-source MCP server. It reads each provider's own model list and tells your assistant which model IDs appeared since it last checked. OpenRouter works without a key; OpenAI, Anthropic, Google Gemini, Mistral, xAI and DeepSeek work with your own. Listing models is free on every one of those APIs, so a check spends no tokens.

Version 1.1.0 · MIT license · no dependencies · Node 18 or newer · source on GitHub

Add it to your client

Claude Code

claude mcp add ai-model-watch -- npx -y github:markksantos/ai-model-watch --mcp

Claude Desktop, Cursor and other MCP clients

Add this to the client's MCP config (claude_desktop_config.json, ~/.cursor/mcp.json and so on):

{
  "mcpServers": {
    "ai-model-watch": {
      "command": "npx",
      "args": ["-y", "github:markksantos/ai-model-watch", "--mcp"]
    }
  }
}

That works as is with OpenRouter. To include another provider, add an env object to the same entry with its key: OpenAI OPENAI_API_KEY, Anthropic ANTHROPIC_API_KEY, Google Gemini GEMINI_API_KEY, Mistral MISTRAL_API_KEY, xAI XAI_API_KEY, DeepSeek DEEPSEEK_API_KEY. Keys go only to their own provider and are never printed.

The two tools

check_new_models: Which model IDs each provider started listing since the last check. The first call records a baseline, so changes show up from the second. Optional providers narrows it to some of openrouter, openai, anthropic, gemini, mistral, xai and deepseek. The lists are saved to ~/.cache/ai-model-watch/state.json, the same file the command-line tool uses.
list_models: What one provider lists right now, filtered by a substring (search) and capped (limit) if you like. Read-only: it doesn't touch the saved lists.

Then just ask: “Any new models on OpenRouter or Anthropic since yesterday?” A new model comes back like this (OpenRouter's two additions on Oct 17, 2026):

OpenRouter: 2 new
  + deepseek/deepseek-v4.1-flash-1017 (DeepSeek: DeepSeek V4.1 Flash 1017), created 2026-10-17
  + reflection/beam-1.0-501b-a23b (Reflection: Beam 1.0 501B A23B), created 2026-10-16

The server, or the radar

The server checks when your assistant calls it, and only those seven APIs. ArtificialWatch runs the same kind of read every 60 seconds across the provider APIs and the public model catalogs, confirms a new model on a second sweep, and then tells you: a free Chrome push the moment it answers, an email about 15 minutes later, texts or a phone call on paid plans. Every call it has made, misses included, is on the track record.

One limit applies to both: a model ID in a list means the provider lists it. It doesn't prove the model answers requests yet, and created is whatever date the provider reports.

New AI model releases · Model status · ai-model-watch on GitHub

sweeping every 60 seconds

Or let the radar watch for you. Free.

A free Chrome push the moment a new model answers on a public API, an email about 15 minutes later. Texts or a phone call on paid plans.

Free forever
No card
Unsubscribe in one click