Aws Bedrock Logs MCP Server

🚀 AWS Bedrock Log Analysis Tool (via Anthropic's MCP Protocol)

This is a command-line interface and API for analyzing the usage and logs of Amazon Web Services (AWS) Bedrock through Anthropic's Model Control Protocol (MCP).

📋 Overview

This tool provides a convenient way to analyze AWS Bedrock model invocation logs and uses Anthropic's Claude model as an interactive interface. It acts as an MCP server, exposing the functionality of the AWS CloudWatch Logs API to Claude, enabling you to query and analyze your Bedrock usage data through natural language.

✨ Features

• Model Usage Analysis: View detailed statistics on Bedrock model usage and token consumption.
• User Analysis: Analyze usage patterns and costs by user.
• Daily Usage Reports: Track daily usage trends and model invocations.
• Token Consumption Metrics: Monitor the usage of input, completion, and total tokens.
• Interactive Interface: Query Bedrock usage data through natural language and interact using Claude.

📋 Requirements

• Python version 3.13 or higher.
• AWS credentials with CloudWatch Logs access.
• Anthropic API access (for Claude integration).

📦 Installation

1. Install uv:

   # On macOS and Linux
   curl -LsSf https://astral.sh/uv/install.sh | sh
   

   # On Windows
   powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
   

2. Clone this repository:

   git clone https://github.com/dheerajoruganty/aws-bedrock-logs-mcp-server.git
   cd aws-bedrock-logs-mcp-server
   

3. Set up a Python virtual environment and install dependencies:

   uv venv && source .venv/bin/activate && uv pip sync pyproject.toml
   

   For Windows:

   uv venv && .venv\Scripts\activate && uv pip sync pyproject.toml
   

4. Configure your AWS credentials:

   mkdir -p ~/.aws
   # Set your credentials in ~/.aws/credentials and ~/.aws/config
   

💻 Usage Examples

🚀 Start the Server

Run the following command to start the server:

python cloudwatch_mcp_server.py

By default, the server communicates with the MCP client using stdio transport.

⚙️ Claude Desktop Configuration

Configure this tool in Claude Desktop:

{
  "mcpServers": {
    "aws_bedrock_logs": {
      "command": "uv",
      "args": [
          "--directory",
          "/path/to/aws-bedrock-logs-mcp",
          "run",
          "cloudwatch_mcp_server.py"
      ]
    }
  }
}

Make sure to replace the directory path with the actual path on your system.

🛠️ Available Tools

The server exposes the following tools to Claude for use:

1. get_bedrock_logs_df: Retrieve the raw Bedrock invocation logs as a pandas DataFrame.
2. get_model_usage_stats: Get usage statistics by model.
3. get_user_usage_stats: Get usage statistics by user.
4. get_daily_usage_stats: Get daily usage statistics and trends.

📝 Example Queries

Once connected to Claude through an MCP-supported interface, you can ask the following questions:

• "Show the Bedrock usage statistics for the past 7 days."
• "What is the average token consumption of each model?"
• "Who are the main Bedrock users ranked by the total number of tokens?"
• "Provide me with a daily breakdown of model invocations."

🔧 Technical Details

📁 Project Structure

• cloudwatch_mcp_server.py: The main server code implementing the MCP tool.
• pyproject.toml: Project dependencies and metadata.
• Dockerfile: Container definition for deployment.

📦 Dependencies

Key dependencies include:

• boto3: A Python SDK for interacting with AWS.
• mcp[cli]: Anthropic's Model Control Protocol library.
• pandas: Data manipulation and analysis.
• pydantic: Data validation using Python type annotations.

📄 License

MIT License

🙏 Acknowledgments

• This tool uses Anthropic's MCP framework.
• It is built on the AWS CloudWatch Logs API.
• The server-side is implemented using FastMCP.