MCP Telemetry
by xprilion
MCP Telemetry provides a simple interface for logging and tracking conversations between users and LLMs. It leverages the Model Context Protocol to expose telemetry tools that can be used to trace and analyze conversations.
Last updated: N/A
What is MCP Telemetry?
MCP Telemetry is a Model Context Protocol (MCP) server designed for telemetry within chat systems, specifically using Weights & Biases Weave. It provides a way to log and track conversations between users and LLMs.
How to use MCP Telemetry?
To use MCP Telemetry, you need to install it by adding the provided JSON configuration to your Claude desktop config. Ensure you have a WandB API key. Once installed, it automatically starts when you launch Claude and collects telemetry data for all conversations. You can then view the data in the Weights & Biases dashboard.
Key features of MCP Telemetry
Start tracing sessions with custom identifiers
Log comprehensive conversation data (user inputs, LLM responses, actions, tool calls)
Seamless integration with Weights & Biases Weave
Real-time monitoring of conversation flows
Use cases of MCP Telemetry
Analyzing conversation flows between users and LLMs
Tracking tool usage patterns in LLM interactions
Monitoring response times of LLMs
Identifying error rates in conversations
FAQ from MCP Telemetry
What is the Model Context Protocol?
What is the Model Context Protocol?
The Model Context Protocol (MCP) is a standard for exposing telemetry tools to trace and analyze conversations.
How do I get a WandB API key?
How do I get a WandB API key?
You can obtain a WandB API key from https://wandb.ai/settings#api.
Where can I view the telemetry data?
Where can I view the telemetry data?
Telemetry data can be viewed in the Weights & Biases dashboard after logging in to your account.
How do I start a tracing session?
How do I start a tracing session?
Prompt Claude to trace the conversation. For example: Log this conversation with MCP Telemetry, topic will be Cats.
What kind of data is logged?
What kind of data is logged?
The server automatically tracks user messages, LLM responses, tool calls and their results, and conversation metadata.