Netdata MCP
An MCP server built into the Netdata agent: metrics, logs, alerts and live process, service and container data for an AI assistant
Install
claude mcp add netdata /usr/bin/nd-mcp --bearer YOUR_API_KEY ws://IP_OF_YOUR_NETDATA:19999/mcpThe key is in /var/lib/netdata/mcp_dev_preview_api_key or /opt/netdata/var/lib/netdata/mcp_dev_preview_api_key.
This is third-party code. Review the repository files before installing.
What it does
Netdata is a monitoring system, and recent versions ship an MCP server inside every agent. Through it the assistant gets node information, discovers and reads system, application and hardware metrics, and views alerts, logs and live data about processes, services, containers, VMs and network connections. Visibility depends on the connection point: a parent node exposes all its children, a single node only itself. Clients with HTTP or SSE support connect to the /mcp endpoint directly, and the nd-mcp bridge that ships with Netdata serves stdio clients.
Who it is for. For DevOps engineers and sysadmins whose infrastructure is already monitored by Netdata.
Good fit when
- You want to find out why a server is slow by asking the assistant instead of browsing charts
- You need to quickly find a node or container with abnormal load
- You need to correlate alerts with logs and processes
Not a fit when
- Netdata is not installed and monitoring runs on another system
- The assistant must not access production logs and process lists
Example request
Check Netdata for which containers used the most CPU in the last hour and whether any alerts are activeLimitations
Requires a recent Netdata installation. Sensitive functions such as logs and live data are protected by a key Netdata generates at startup. Connecting through Netdata Cloud is not available yet, so you connect to a specific agent or parent node. The assistant's model may need a VPN, but the Netdata server runs on your side.
How to disable. Remove the netdata server from the client's MCP config or run claude mcp remove netdata.
MCP
- Transport
- http
- Authentication
- not required
Security check
- The assistant gains access to server logs, processes and network connections
- The access key must be protected like a password
README in short
The MCP docs in the repository explain how the assistant's visibility depends on where it connects in the Netdata hierarchy. They give a shared JSON config with the nd-mcp bridge, bridge paths for different installs, direct HTTP and SSE connections and where the access key lives. Claude Desktop, Claude Code and Cursor are covered separately, along with alternative Go, Node.js and Python bridges and a capability list.
FAQ
Is a separate install needed?
No, the server is built into Netdata and the nd-mcp bridge ships with every install. Its path depends on how Netdata was installed.
Can I skip the bridge?
Yes, if the client supports HTTP or SSE: use http://IP:19999/mcp with an Authorization Bearer header.
Related
GitHub's official MCP server: code, issues, pull requests, Actions and security alerts straight from the agent
AWS's official MCP server suite: docs, IaC, containers, serverless, databases, cost and monitoring
A GitHub Action that runs Claude Code in pull requests and issues: answers mentions, reviews and makes changes
An AI assistant for Kubernetes that turns plain-language requests into kubectl operations and doubles as an MCP server