Agent♥︎Age
Catalog

io.github.bch1212/modelwatch

Official

by bch1212 · Python

Continuous behavioral drift monitoring for LLM apps — catches silent provider model updates.

modelwatch-mcp (Model Context Protocol) Server

modelwatch-mcp is an MCP server for ModelWatch that performs continuous behavioral drift monitoring for LLM-powered applications. It enables an MCP-compatible client (including Claude Desktop or Claude Code) to create drift specs, run them, and surface drift events within the chat workflow.

🛠️ Key Features

  • Continuous behavioral drift monitoring for LLM apps
  • Create drift specs from an MCP-compatible client
  • Run drift specs and surface drift events

🚀 Use Cases

  • Detect silent provider model updates in LLM applications
  • Run drift checks from chat-based tooling (Claude Desktop/Claude Code)

⚡ Developer Benefits

  • Integrates with the Model Context Protocol via an MCP server
  • Supports drift-spec execution without leaving the chat

⚠️ Limitations

  • Configuration details are shown for Claude Desktop; other client configuration is not included in the provided data

Topics

ai-agentsanthropicclaudedrift-detectionllmllm-monitoringmcpmcp-servermodel-context-protocolobservabilitytypescript