Part of Grafana, not a vendor of its own. The four checks, setup and context cost live in the Grafana guide.

Part of the Grafana stack

Grafana Mimir and MCP

What it is

Grafana's metrics backend, a horizontally scalable store for Prometheus metrics. It speaks PromQL, so nothing in the connector is Mimir-specific.

How an agent reaches it

The main Grafana MCP server, through its six PromQL tools. Those same tools serve Prometheus, Thanos, Mimir and Cloud Monitoring, so the datasource decides which backend answers.

The tools that touch it

list_prometheus_metric_names

Discovery. The tool description says to call this first, before querying anything.

list_prometheus_label_names

Lists label names in a PromQL-compatible datasource.

list_prometheus_label_values

Lists the values of one label, to filter a query down.

list_prometheus_metric_metadata

Metadata about the metrics currently scraped from targets.

query_prometheus

Runs the PromQL query. Its description spells out the workflow: names, then label values, then query.

query_prometheus_histogram

Histogram percentiles, with the same discover-first instruction.

Six tools, all on by default, counted from the captured tools/list for the Grafana MCP server. None mentions Mimir by name: they target any PromQL-compatible datasource, and Mimir is one.

The limit an agent runs into

Mimir allows 2 minutes for a query, but the MCP server gives up after 10 seconds by default, which is the shortest timeout in the path. The backend may have been about to answer; the connector stopped waiting.

What to do: Raise the connector's timeout rather than the backend's. The 10-second default is the binding constraint, so tuning Mimir changes nothing until that is lifted.

Reference

Tools

6, all on by default

Query language

PromQL

Same tools also serve

Prometheus, Thanos and Cloud Monitoring

Mimir query allowance

2 minutes

Connector default timeout

10 seconds

Binding constraint

The connector, not the backend