ModelRot

MCP server

Give your coding agent today's model lifecycle

Agents write model IDs they learned months ago. Add ModelRot and they can check a model, find its replacement and scan a file before they commit. Free for 100 calls a day without signing up.

claude mcp add --transport http modelrot https://modelrot.vibiz.app/mcp

Server URL https://modelrot.vibiz.app/mcp · streamable HTTP · no sign-in needed. Create a free key for 1,000 calls a day (Pro 5,000, Team 50,000).

Tools

check_model(id)

Check whether an LLM model ID (OpenAI, Anthropic, Google Gemini, Mistral, xAI or Amazon Bedrock) is active, legacy, deprecated or retired, when it shuts off, what replaces it and what it costs. Call this before writing a model ID into code.

list_retiring(provider, within_days)

List model IDs that shut off within the next N days, optionally for one provider, soonest first.

suggest_replacement(id, priority)

Suggest what to use instead of a model ID: the provider's documented replacement first, then current alternatives from the same provider, ordered by cost (cheapest first) or by quality (the provider's higher tiers first).

get_price(id)

Get the standard input and output price per 1M tokens for a model ID, from the provider’s pricing page.

scan_code(code)

Find every LLM model ID in a code snippet (up to 200 KB) and report which ones are retiring, deprecated or retired.

Tell your agent to use it

Add this to CLAUDE.md, AGENTS.md or .cursorrules so the agent checks before writing a model ID:

Before adding or changing an LLM model ID in code, call the modelrot
check_model tool. If it is deprecated or retired, use the replacement
from suggest_replacement instead. Run scan_code on files that call LLMs.

Try it with curl

curl -s https://modelrot.vibiz.app/mcp \
  -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call",
       "params":{"name":"check_model","arguments":{"id":"gpt-4o-2024-08-06"}}}'