newrelic

This MCP server provides programmatic access to New Relic data via a REST-like MCP interface, enabling AI assistants and other clients to interact with APM, NRQL, logs, alerts, dashboards, and more. It exposes a collection of tools such as application management (list/get applications and their metrics), NRQL and NerdGraph querying, log parsing rule management, alert policies, synthetic monitors, dashboards, entity search, infrastructure data, deployments, and user management. You can query NRQL, run NerdGraph queries, manage log parsing GROK patterns, and orchestrate New Relic resources from your MCP-enabled assistant or workflow. To use it, run the server and connect through your MCP client configuration, or invoke it directly if you have the Python module installed, as shown in the examples.

Prerequisites:

• Python 3.8 or newer
• pip (Python package installer)

Option A: Install from PyPI (Recommended)

pip install newrelic-mcp-server

Option B: Install from Source

# Clone this repository
git clone https://github.com/piekstra/newrelic-mcp-server.git
cd newrelic-mcp-server

# Install in development mode
pip install -e .

Running the server:

• After installation, you can start the MCP server with:

newrelic-mcp-server

Or, if using the module approach:

python -m newrelic_mcp

Environment and credentials:

• Prefer using macOS Keychain storage for credentials (secure setup) via: newrelic-mcp-setup or python -m newrelic_mcp.credentials
• If using environment variables instead, set NEWRELIC_API_KEY, NEWRELIC_REGION (US by default), and NEWRELIC_ACCOUNT_ID as needed. The server will prefer keychain credentials when both are available.

Common issues:

• Ensure your API key has the necessary permissions (New Relic User API key starting with NRAK).
• If you encounter connection issues, verify that the environment variables are correctly set or that the keychain is accessible.
• When running behind proxies or restricted networks, ensure outbound access to New Relic endpoints is allowed.

Configuration tips:

• In Claude Desktop or other MCP clients, you can specify the server command and environment variables as shown in the README to securely inject credentials.
• The server supports a broad range of tools; you can selectively enable or query by using the appropriate method names (e.g., list_applications, query_nrql, list_dashboards, etc.).