Mackerel MCP Server
Official# Mackerel MCP Server
A Model Context Protocol server for interacting with [Mackerel](https://mackerel.io).
# Configuration
The MCP server can be run either via Docker or npx.
## Running via Docker
```json
{
"mcpServers": {
"mackerel": {
"command": "docker",
"args": [
"run",
"--rm",
"-i",
"-e",
"MACKEREL_APIKEY",
"ghcr.io/mackerelio-labs/mcp-server:latest"
],
"env": {
"MACKEREL_APIKEY": "${MACKEREL_APIKEY}"
}
}
}
}
```
## Running via npx
```json
{
"mcpServers": {
"mackerel": {
"command": "npx",
"args": [
"-y",
"@mackerel/mcp-server"
],
"env": {
"MACKEREL_APIKEY": "${MACKEREL_APIKEY}"
}
}
}
}
```
# Available tools
* `list_alerts` - Retrieve alerts.
* `get_alert` - Retrieve a specific alert.
* `get_alert_logs` - Retrieve logs for a specific alert.
* `list_dashboards` - Retrieve dashboards.
* `get_dashboard` - Retrieve a specific dashboard.
* `update_dashboard` - Update a specific dashboard.
* `list_hosts` - Retrieve hosts.
* `get_host_metrics` - Retrieve metrics data for a specific host.
* `list_services` - Retrieve services.
* `get_service_metrics` - Retrieve metrics data for a specific service.
* `list_monitors` - Retrieve monitor configurations.
* `get_monitor` - Retrieve a specific monitor configuration.
* `get_trace` - Retrieve trace data by trace ID for distributed tracing analysis.
* `list_traces`: - Search and retrieve traces.
* `list_http_server_stats` - Retrieve HTTP server statistics.
* `list_db_query_stats` - Retrieve database query statistics.
* `search_logs` - Search logs of a service.
TDQS
Scored across 17 tools
Each tool targets a clearly distinct resource and action: alerts, dashboards, hosts, services, monitors, metrics, traces, HTTP stats, DB stats, and logs. Overlaps such as get_trace vs. search_logs or get_host_metrics vs. get_service_metrics are resolved by resource scope and descriptions.
All tool names follow a consistent snake_case verb_noun convention, with list_*, get_*, update_*, and search_* used predictably. The one search_* tool still fits the same readable pattern.
Seventeen tools is slightly above the typical 3-15 range, but the server covers a broad monitoring domain with many distinct resource types. Each tool appears to earn its place, though the surface is on the heavier side.
The read-oriented surface is broad, but lifecycle coverage is incomplete: dashboards have list/get/update but no create/delete, monitors have list/get but no create/update/delete, and hosts/services lack dedicated detail tools. These gaps are workable for monitoring queries but notable for administrative workflows.