Metrics

The MCP server supports querying Prometheus-compatible metrics using PromQL. Use these tools to discover metrics, explore metric labels, and query time-series data to monitor service behavior in real time.

Search metrics

Use the search_relevant_metrics tool to find metrics before writing PromQL queries. Search semantically by natural-language description, or match metric names directly with a wildcard name pattern.
Name Required Type Description
query Yes string Search text. Its meaning depends on mode.
mode No enum by_description (semantic search, default) or by_name (wildcard name pattern match, e.g. http_*, *_total).
top_k No number Maximum number of semantic matches to return (1–100, default: 5). Ignored when mode is by_name.

Explore metric labels

Use the get_metric_labels tool to list the label names available for a metric, then use get_metric_label_values to list the values for a specific label. Use these to understand a metric's structure before constructing queries.

get_metric_labels parameters:
Name Required Type Description
metric_name Yes string The metric to list label names for.

get_metric_label_values parameters:
Name Required Type Description
metric_name Yes string The metric to list label values for.
label Yes string The label to list values for.

Instant metric query

Run a PromQL expression at a single point in time using the query_promql_instant tool.

Example:

{
  "query": "sum by (service) (rate(http_requests_total[5m]))",
  "time": "2024-01-15T11:00:00Z"
}

Parameters:
Name Required Type Description
query Yes string A valid PromQL expression evaluated at a single point in time.
time No string RFC-3339 UTC timestamp for the evaluation time. Defaults to the current time.

Range metric query

Run a PromQL expression across a time window using the query_promql_range tool.

Example:

{
  "query": "cpu_usage_percent",
  "start": "2024-01-15T10:00:00Z",
  "end": "2024-01-15T11:00:00Z",
  "step": "1m",
  "limit": 20
}

Parameters:
Name Description
query A valid PromQL expression to evaluate over a range of time. Aggregate and segment with by(label), and avoid high-cardinality results.
start Start time of the query window (RFC-3339 UTC format).
end End time of the query window (RFC-3339 UTC format).
step Interval between data points, specified as a duration (e.g., 1m, 30s, 6h, 1d). Size it so each series returns roughly 20 data points.
limit Maximum number of samples to keep per time-series (1–200). Series that exceed the limit are marked as truncated.

Need help? Contact Support.

What's new? Find out here.

LLM? Read llms.txt.

Previous Logs and traces

Next Real User Monitoring

Name	Required	Type	Description
`query`	Yes	string	Search text. Its meaning depends on `mode`.
`mode`	No	enum	`by_description` (semantic search, default) or `by_name` (wildcard name pattern match, e.g. `http_`, `_total`).
`top_k`	No	number	Maximum number of semantic matches to return (1–100, default: 5). Ignored when `mode` is `by_name`.

Name	Required	Type	Description
`metric_name`	Yes	string	The metric to list label values for.
`label`	Yes	string	The label to list values for.

Name	Required	Type	Description
`query`	Yes	string	A valid PromQL expression evaluated at a single point in time.
`time`	No	string	RFC-3339 UTC timestamp for the evaluation time. Defaults to the current time.

Name	Description
`query`	A valid PromQL expression to evaluate over a range of time. Aggregate and segment with `by(label)`, and avoid high-cardinality results.
`start`	Start time of the query window (RFC-3339 UTC format).
`end`	End time of the query window (RFC-3339 UTC format).
`step`	Interval between data points, specified as a duration (e.g., `1m`, `30s`, `6h`, `1d`). Size it so each series returns roughly 20 data points.
`limit`	Maximum number of samples to keep per time-series (1–200). Series that exceed the limit are marked as truncated.