Bring infrastructure monitoring into your AI workflow.

Connect Copilot, Claude Code, or Cursor directly to Hosted Graphite to search metrics, render data, manage alerts, and configure notifications using natural language.

Connect the MCP server See how it works

14-day free trial. No credit card. No dashboard required to get started.

How it works

Infrastructure monitoring, directly from your AI agent

Connect Claude Code, Cursor, or GitHub Copilot to Hosted Graphite to search metrics, render data, manage alerts, and configure notifications using natural language.

Point your app metrics to MetricFire

Send app metrics from wherever your app already runs, over HTTP POST, TCP, UDP, StatsD or an agent.

Connect the MCP server to your AI agent

Claude Code, Cursor, and GitHub Copilot can all talk to Hosted Graphite directly. One token, pasted into your agent's MCP config, and it can see your metrics.

Ask for the monitoring you actually want

Graphite alerting is built in, with optional composite AND/OR logic. Ask your agent to find metrics, render graphs, build alerts, wire up Slack, and schedule mutes, all in plain language.

Video Tutorial

Connect the MCP Server in under five minutes

Sign up and create a token. In your MetricFire account, go to Addons, then MetricFire MCP Server, and generate a token scoped to what your agent needs.
Point your app metrics to MetricFire. Send app metrics from wherever your app already runs, over HTTP POST, TCP, UDP, StatsD or an agent.
Add the server to your agent's MCP config. Works with Claude Code, Cursor, and GitHub Copilot.
Ask. "List all available Hosted Graphite tools" is a good place to start.

Learn More

Copilot / VS Code configuration example

{
  "servers": {
    "mf-mcp-server": {
      "url": "https://mcp.hostedgraphite.com/",
      "headers": {
        "X-HostedGraphite-MCP-Token": "MF_MCP_TOKEN"
      }
    }
  }
}

Config shape varies slightly by client — see the docs for Claude, Cursor, Cline, Windsurf, and others.

Just ask

Things vibe coders actually ask their agent for

Every one of these is a real MCP tool call, the same ones used for metric search, rendering, and alert management. These workflows assume you're already sending metrics into your Hosted Graphite account. This is how the MCP tool can use your existing data to bring infrastructure monitoring into your AI workflow.

"What metrics do I even have right now?"

Searches your metric namespace, no memorized paths needed.

"Show me API latency over the last hour."

Renders the data as a chart, right in the chat.

"Tell me if error rate goes over 5% for 5 minutes."

Builds the alert condition and saves it, no form to fill in.

"Post that to my Slack when it fires."

Attaches an existing notification channel to the alert.

"Mute this on weekends, nobody's shipping then."

Creates a scheduled mute for exactly the window you name.

"Combine three conditions so it only fires once, for real."

Composite AND/OR logic, so one flaky metric doesn't spam you.

$19 /month to start

One metric is one metric. No per-user fees, no cardinality charges.

14-day trialNo card required

Graphite alertingBuilt in on every plan

Loki loggingAvailable alongside metrics

No lock-inFull data export

Your app is live. Give it a way to tell you when it's not.

Start free, connect the MCP server to your agent, and ask for the first alert before you close the chat.

Start free trial