AWS Bedrock Logs MCP

Example Use Cases

• 1

  Query and filter Bedrock invocation logs from AWS CloudWatch in natural language to troubleshoot model calls or investigate specific events.

• 2

  Compute and summarize model usage and token consumption (input/completion/total) across models and users for cost analysis and optimization.

• 3

  Daily or time-series reports of Bedrock invocations and trends to monitor activity, detect anomalies, and produce usage summaries for stakeholders.

Description

This project runs an MCP (Model Control Protocol) server that lets Claude (Anthropic) interact with AWS CloudWatch Logs to analyze AWS Bedrock model invocations. It exposes utilities to fetch raw logs as DataFrames, compute model- and user-level usage statistics, and produce daily usage trends and token consumption metrics. The server enables natural-language querying and reporting of Bedrock usage, making it easier to investigate costs, token usage, and user activity. It requires AWS credentials and Anthropic API access to operate.