datadog-mcp-server

datadog-mcp-server

0
databasesmonitoringdeveloper_tools

FastMCP is a lightweight server implementing the Model Context Protocol, designed to manage and query model contexts efficiently. It features integration with Datadog for metrics and fast development using the FastMCP framework.

FastMCP - Model Context Protocol Server

A lightweight Model Context Protocol (MCP) server implemented with FastMCP, a fast and Pythonic framework for building MCP servers and clients.

Features

  • Create, retrieve, update, and delete model contexts
  • Query execution against specific contexts
  • Filtering by model name and tags
  • In-memory storage (for development)
  • FastMCP integration for easy MCP server development
  • Datadog integration for metrics and monitoring

Requirements

  • Python 3.7+
  • FastMCP
  • uv (recommended for environment management)
  • Datadog account (optional, for metrics)

Installation

Using uv (Recommended)

The simplest way to install is using the provided scripts:

Unix/Linux/macOS
# Clone the repository
git clone https://github.com/yourusername/datadog-mcp-server.git
cd datadog-mcp-server

# Make the install script executable
chmod +x install.sh

# Run the installer
./install.sh
Windows
# Clone the repository
git clone https://github.com/yourusername/datadog-mcp-server.git
cd datadog-mcp-server

# Run the installer
.\install.ps1

Manual Installation

# Clone the repository
git clone https://github.com/yourusername/datadog-mcp-server.git
cd datadog-mcp-server

# Create and activate a virtual environment with uv
uv venv
# On Unix/Linux/macOS:
source .venv/bin/activate
# On Windows:
.\.venv\Scripts\activate

# Install dependencies
uv pip install -r requirements.txt

Datadog Configuration

The server integrates with Datadog for metrics and monitoring. You can configure Datadog API credentials in several ways:

1. Environment Variables

Set these environment variables before starting the server:

# Unix/Linux/macOS
export DATADOG_API_KEY=your_api_key
export DATADOG_APP_KEY=your_app_key  # Optional
export DATADOG_SITE=datadoghq.com    # Optional, default: datadoghq.com

# Windows PowerShell
$env:DATADOG_API_KEY = 'your_api_key'
$env:DATADOG_APP_KEY = 'your_app_key'  # Optional
$env:DATADOG_SITE = 'datadoghq.com'    # Optional

2. .env File

Create a .env file in the project directory:

DATADOG_API_KEY=your_api_key
DATADOG_APP_KEY=your_app_key
DATADOG_SITE=datadoghq.com

3. FastMCP CLI Installation

When installing as a Claude Desktop tool, you can pass environment variables:

fastmcp install mcp_server.py --name "Model Context Server" -v DATADOG_API_KEY=your_api_key

4. Runtime Configuration

Use the configure_datadog tool at runtime:

result = await client.call_tool("configure_datadog", {
    "api_key": "your_api_key",
    "app_key": "your_app_key",  # Optional
    "site": "datadoghq.com"     # Optional
})

Usage

Starting the Server

# Start directly from activated environment
python mcp_server.py

# Or use uv run (no activation needed)
uv run python mcp_server.py

# Use FastMCP CLI for development (if in activated environment)
fastmcp dev mcp_server.py

# Use FastMCP CLI with uv (no activation needed)
uv run -m fastmcp dev mcp_server.py

Installing as a Claude Desktop Tool

# From activated environment
fastmcp install mcp_server.py --name "Model Context Server"

# Using uv directly
uv run python -m fastmcp install mcp_server.py --name "Model Context Server"

# With Datadog API key
fastmcp install mcp_server.py --name "Model Context Server" -v DATADOG_API_KEY=your_api_key

Using the Tools

The server provides the following tools:

  • create_context - Create a new context
  • get_context - Retrieve a specific context
  • update_context - Update an existing context
  • delete_context - Delete a context
  • list_contexts - List all contexts (with optional filtering)
  • query_model - Execute a query against a specific context
  • health_check - Server health check
  • configure_datadog - Configure Datadog integration at runtime

Example Requests

Creating a Context

result = await client.call_tool("create_context", {
    "context_id": "model-123",
    "model_name": "gpt-3.5",
    "data": {
        "parameters": {
            "temperature": 0.7
        }
    },
    "tags": ["production", "nlp"]
})

Executing a Query

result = await client.call_tool("query_model", {
    "context_id": "model-123",
    "query_data": {
        "prompt": "Hello, world!"
    }
})

Configuring Datadog

result = await client.call_tool("configure_datadog", {
    "api_key": "your_datadog_api_key",
    "app_key": "your_datadog_app_key",  # Optional
    "site": "datadoghq.com"             # Optional
})

Datadog Metrics

The server reports the following metrics to Datadog:

  • mcp.contexts.created - Context creation events
  • mcp.contexts.updated - Context update events
  • mcp.contexts.deleted - Context deletion events
  • mcp.contexts.accessed - Context access events
  • mcp.contexts.total - Total number of contexts
  • mcp.contexts.listed - List contexts operation events
  • mcp.queries.executed - Query execution events
  • mcp.server.startup - Server startup events
  • mcp.server.shutdown - Server shutdown events

Development

See the included mcp_example.py for a client implementation example:

# Run the example client (with activated environment)
python mcp_example.py

# Run with uv (no activation needed)
uv run python mcp_example.py

License

MIT

Resources