Claude Grafana connectorWhat Claude can do in your Grafana account.

Searches the dashboards of your Grafana instance from a few words. It is often Claude's first call, before opening anything at all.

Lists the data sources registered in your instance: Prometheus, Loki, ClickHouse and the rest. Claude then knows which source to query.

Opens a data source record from its unique ID. Handy when a dashboard or query cites the ID without the name.

Fetches a data source by its readable name, to get its type and ID before querying it.

Loads a complete dashboard from its ID, with its full definition.

Gives a compact summary of a dashboard without loading its full definition, which spares the conversation's context.

Pulls one targeted part of a dashboard through a path expression, without loading the rest.

Returns, for each panel of a dashboard, its title, its query, and the ID and type of its data source.

Runs a PromQL query against a Prometheus source and returns the values. Claude can compose the query from your plain-English question.

Calculates percentiles from a Prometheus histogram, so you do not have to hand-craft the function yourself.

Lists the metadata Prometheus holds on its metrics. Claude gets a better idea of what a metric measures before relying on it.

Returns the metric names available on a Prometheus source.

Returns Prometheus label names matching a selector, such as service, region or environment.

Gives the values a specific Prometheus label takes, for instance every region or every instrumented service.

Queries Loki in LogQL, either to read log lines or to compute a metric from them.

Lists every label name available in Loki logs, so Claude knows how to filter before it runs a log query.

Shows the values of a given Loki log label, for example the list of applications present in the logs.

Provides statistics on Loki log streams rather than the lines themselves. A way to measure before reading.

Surfaces recurring patterns detected in Loki logs, to spot repeated structures without reading everything.

Lists the alert rules configured in your Grafana instance. Claude gets the overall picture before opening a single rule.

Opens one alert rule in full from its ID, once Claude has spotted it in the list.

Lists notification contact points, both those Grafana manages and those of an external Alertmanager. Claude sees where each notification is meant to land.

Lists incidents from Grafana Incident. The Viewer role is enough for this lookup. It gives the overall picture first.

Retrieves one Grafana Incident record by its ID, with its details.

Returns the list of Sift investigations, Grafana's automated error and performance analysis, with an optional result limit. It is the entry point to Sift.

Loads an existing Sift investigation from its unique ID, so Claude can go through it with you.

Pulls one specific analysis inside a Sift investigation, to go deep on a single line of inquiry. It keeps the answer on one angle.

Uses Sift to detect error patterns rising abnormally in Loki logs.

Relies on Sift to spot slow requests in the relevant Tempo trace sources. It works from trace data, not from logs.

Lists on-call schedules from Grafana OnCall. Each rotation shows up separately.

Details one specific Grafana OnCall shift, picked out by its ID. It is the zoom that follows a look at the schedules.

Tells you who is on call right now for a given schedule.

Enumerates the teams declared in Grafana OnCall, to scope an on-call question to the right group.

Lists the users Grafana OnCall knows about, so Claude can match a name to the right person in a rotation.

Lists Grafana OnCall alert groups, with filtering options to narrow the list. It is where Claude starts when you ask what has been firing.

Opens one OnCall alert group from its ID. You get the full record.

Lists every team configured in Grafana. It is an admin tool, switched off by default in the server.

Enumerates the users of a Grafana organization. It belongs to the admin tools, off by default. It is a plain roster.

Lists every role defined in Grafana, so Claude can see what exists before looking at who holds what. Admin tool, off by default in the server.

Details one Grafana role, so you can see what sits behind its name.

Lists the assignments of a given role. It answers the question of who, not what.

Lists the roles assigned to one or more Grafana users.

For one or more Grafana teams, returns the roles they hold.

Lists the permissions defined on one Grafana resource. Claude sees who can do what on it without clicking through the settings.

Describes a type of Grafana resource, as seen from the permissions side.

Gives Pyroscope label names matching a selector. Pyroscope is Grafana's continuous profiling tool. It is where profiling questions start.

Returns the values of a Pyroscope label for a given selector, for instance the services being profiled.

Lists the profile types available in Pyroscope. It frames what you can ask next about an application's performance.

Fetches a Pyroscope profile in DOT format, a graph format Claude can then read to spot costly functions.

Builds exact links to Grafana resources, so Claude does not have to guess URLs.

Retrieves annotations, the markers placed on graphs, with filters to target the ones you need.

Lists the tags used on annotations, with an optional filter. It is a quick sanity check.

Renders a PNG image of a single panel or of a whole dashboard.

Returns an assertion summary for a given entity, with permissions specific to the plugin involved.

Searches logs across ClickHouse and Loki at once, in a single request. Grafana switches it off by default in its server.

Fires a SQL query at a ClickHouse source, with Grafana macro substitution. Off by default in the server.

Lists the tables in a ClickHouse database. It belongs to the ClickHouse family, disabled by default.

Gives a ClickHouse table's schema, with each column's type. Claude can then get a query right the first time.

Provides sample queries for a data source type, to help Claude phrase a valid query.