Grafana MCP server

Name: Grafana MCP server
Author: Grafana Labs

Search and edit Grafana dashboards, query metrics and logs, and handle alerts

Official · Grafana Labs Go Remote (hosted) · Streamable HTTP Local · stdio DevOps and monitoring Data and analytics

GitHub PyPI Docker Hub Website

103 popularity score ★ 3.6k ↓ 740k/mo

Connect your client

Remote (hosted)

Add in one click

Add to Cursor Add to VS Code Add to LM Studio Add to Goose

Opens the app, which asks before it adds the server. Then replace each <NAME> in its settings with your value.

Or add it by command or config

Cursor.cursor/mcp.json

{
  "mcpServers": {
    "grafana": {
      "url": "https://mcp.grafana.com/mcp",
      "headers": {
        "X-Grafana-URL": "https://<YOUR_STACK>.grafana.net"
      }
    }
  }
}

Claude Desktopclaude_desktop_config.json

{
  "mcpServers": {
    "grafana": {
      "type": "http",
      "url": "https://mcp.grafana.com/mcp",
      "headers": {
        "X-Grafana-URL": "https://<YOUR_STACK>.grafana.net"
      }
    }
  }
}

Claude Codeterminal

claude mcp add grafana https://mcp.grafana.com/mcp \
  --transport http \
  --header "X-Grafana-URL: https://<YOUR_STACK>.grafana.net"

windsurf~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "grafana": {
      "url": "https://mcp.grafana.com/mcp",
      "headers": {
        "X-Grafana-URL": "https://<YOUR_STACK>.grafana.net"
      }
    }
  }
}

VS Code.vscode/mcp.json

{
  "servers": {
    "grafana": {
      "type": "http",
      "url": "https://mcp.grafana.com/mcp",
      "headers": {
        "X-Grafana-URL": "https://<YOUR_STACK>.grafana.net"
      }
    }
  }
}

Local (PyPI)

Add in one click

Add to Cursor Add to VS Code Add to LM Studio Add to Goose

Opens the app, which asks before it adds the server. Then replace each <NAME> in its settings with your value.

Or add it by command or config

It needs GRAFANA_URL, GRAFANA_SERVICE_ACCOUNT_TOKEN. Replace each <NAME> with your value.

Claude CodeTerminal

claude mcp add grafana -e GRAFANA_URL=<GRAFANA_URL> -e GRAFANA_SERVICE_ACCOUNT_TOKEN=<GRAFANA_SERVICE_ACCOUNT_TOKEN> -- uvx mcp-grafana

Claude Desktopclaude_desktop_config.json

{
  "mcpServers": {
    "grafana": {
      "command": "uvx",
      "args": [
        "mcp-grafana"
      ],
      "env": {
        "GRAFANA_URL": "<GRAFANA_URL>",
        "GRAFANA_SERVICE_ACCOUNT_TOKEN": "<GRAFANA_SERVICE_ACCOUNT_TOKEN>"
      }
    }
  }
}

Cursor~/.cursor/mcp.json

{
  "mcpServers": {
    "grafana": {
      "command": "uvx",
      "args": [
        "mcp-grafana"
      ],
      "env": {
        "GRAFANA_URL": "<GRAFANA_URL>",
        "GRAFANA_SERVICE_ACCOUNT_TOKEN": "<GRAFANA_SERVICE_ACCOUNT_TOKEN>"
      }
    }
  }
}

VS Code.vscode/mcp.json

{
  "servers": {
    "grafana": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "mcp-grafana"
      ],
      "env": {
        "GRAFANA_URL": "<GRAFANA_URL>",
        "GRAFANA_SERVICE_ACCOUNT_TOKEN": "<GRAFANA_SERVICE_ACCOUNT_TOKEN>"
      }
    }
  }
}

Codex~/.codex/config.toml

[mcp_servers.grafana]
command = "uvx"
args = ["mcp-grafana"]

[mcp_servers.grafana.env]
GRAFANA_URL = "<GRAFANA_URL>"
GRAFANA_SERVICE_ACCOUNT_TOKEN = "<GRAFANA_SERVICE_ACCOUNT_TOKEN>"

Gemini CLITerminal

gemini mcp add --scope user -e GRAFANA_URL=<GRAFANA_URL> -e GRAFANA_SERVICE_ACCOUNT_TOKEN=<GRAFANA_SERVICE_ACCOUNT_TOKEN> grafana uvx -- mcp-grafana

Local (Docker)

Add in one click

Add to Cursor Add to VS Code Add to LM Studio Add to Goose

Opens the app, which asks before it adds the server. Then replace each <NAME> in its settings with your value.

Or add it by command or config

It needs GRAFANA_URL, GRAFANA_SERVICE_ACCOUNT_TOKEN. Replace each <NAME> with your value.

Claude CodeTerminal

claude mcp add grafana -e GRAFANA_URL=<GRAFANA_URL> -e GRAFANA_SERVICE_ACCOUNT_TOKEN=<GRAFANA_SERVICE_ACCOUNT_TOKEN> -- docker run --rm -i -e GRAFANA_URL -e GRAFANA_SERVICE_ACCOUNT_TOKEN grafana/mcp-grafana -t stdio

Claude Desktopclaude_desktop_config.json

{
  "mcpServers": {
    "grafana": {
      "command": "docker",
      "args": [
        "run",
        "--rm",
        "-i",
        "-e",
        "GRAFANA_URL",
        "-e",
        "GRAFANA_SERVICE_ACCOUNT_TOKEN",
        "grafana/mcp-grafana",
        "-t",
        "stdio"
      ],
      "env": {
        "GRAFANA_URL": "<GRAFANA_URL>",
        "GRAFANA_SERVICE_ACCOUNT_TOKEN": "<GRAFANA_SERVICE_ACCOUNT_TOKEN>"
      }
    }
  }
}

Cursor~/.cursor/mcp.json

{
  "mcpServers": {
    "grafana": {
      "command": "docker",
      "args": [
        "run",
        "--rm",
        "-i",
        "-e",
        "GRAFANA_URL",
        "-e",
        "GRAFANA_SERVICE_ACCOUNT_TOKEN",
        "grafana/mcp-grafana",
        "-t",
        "stdio"
      ],
      "env": {
        "GRAFANA_URL": "<GRAFANA_URL>",
        "GRAFANA_SERVICE_ACCOUNT_TOKEN": "<GRAFANA_SERVICE_ACCOUNT_TOKEN>"
      }
    }
  }
}

VS Code.vscode/mcp.json

{
  "servers": {
    "grafana": {
      "type": "stdio",
      "command": "docker",
      "args": [
        "run",
        "--rm",
        "-i",
        "-e",
        "GRAFANA_URL",
        "-e",
        "GRAFANA_SERVICE_ACCOUNT_TOKEN",
        "grafana/mcp-grafana",
        "-t",
        "stdio"
      ],
      "env": {
        "GRAFANA_URL": "<GRAFANA_URL>",
        "GRAFANA_SERVICE_ACCOUNT_TOKEN": "<GRAFANA_SERVICE_ACCOUNT_TOKEN>"
      }
    }
  }
}

Codex~/.codex/config.toml

[mcp_servers.grafana]
command = "docker"
args = ["run", "--rm", "-i", "-e", "GRAFANA_URL", "-e", "GRAFANA_SERVICE_ACCOUNT_TOKEN", "grafana/mcp-grafana", "-t", "stdio"]

[mcp_servers.grafana.env]
GRAFANA_URL = "<GRAFANA_URL>"
GRAFANA_SERVICE_ACCOUNT_TOKEN = "<GRAFANA_SERVICE_ACCOUNT_TOKEN>"

Gemini CLITerminal

gemini mcp add --scope user -e GRAFANA_URL=<GRAFANA_URL> -e GRAFANA_SERVICE_ACCOUNT_TOKEN=<GRAFANA_SERVICE_ACCOUNT_TOKEN> grafana docker -- run --rm -i -e GRAFANA_URL -e GRAFANA_SERVICE_ACCOUNT_TOKEN grafana/mcp-grafana -t stdio

Settings 12

Set together: prometheus guardrail
`GRAFANA_URL`	Required: Address of your Grafana instance or Grafana Cloud stack. Default `http://localhost:3000`.
`GRAFANA_SERVICE_ACCOUNT_TOKEN` secret	Required: Service account token. GRAFANA_SERVICE_ACCOUNT_TOKEN_FILE or GRAFANA_USERNAME with GRAFANA_PASSWORD can be used instead.
`GRAFANA_SERVICE_ACCOUNT_TOKEN_FILE`	Optional: Path to a file holding the token, read again on each request so it can be rotated.
`GRAFANA_ORG_ID`	Optional: Numeric organization ID on multi-org instances.
`GRAFANA_EXTRA_HEADERS`	Optional: Extra HTTP headers sent with every request to Grafana, as JSON.
`GRAFANA_USAGE_STATS`	Optional: Set to disabled to stop usage reports, or log to print them instead. DO_NOT_TRACK=1 also turns them off. Default `enabled`.
`MCP_GRAFANA_SERVER_TOKEN` secret	Optional: Bearer token clients must send when the server runs over HTTP. For SSE and streamable HTTP transports.
`GRAFANA_PROMETHEUS_GUARDRAIL_MODE`	Optional: off, shadow (log what would be blocked) or enforce (reject queries over the caps below). Default `off`. For query_prometheus, query_prometheus_histogram, run_panel_query.
`GRAFANA_PROMETHEUS_GUARDRAIL_MAX_RANGE`	Optional: Longest time range one PromQL query may read, counting range vectors and subqueries. 0 turns the check off. Default `192h`.
`GRAFANA_PROMETHEUS_GUARDRAIL_MAX_POINTS`	Optional: Most points per series a range query may ask for. 0 turns the check off. Default `11000`.
Set together: loki guardrail
`GRAFANA_LOKI_GUARDRAIL_MODE`	Optional: off, shadow or enforce. GRAFANA_LOKI_GUARDRAIL_MAX_RANGE (24h by default) and GRAFANA_LOKI_GUARDRAIL_MAX_BYTES set its caps. Default `off`. For query_loki_logs, query_loki_patterns.
`GRAFANA_LOKI_GUARDRAIL_ON_ESTIMATE_FAILURE`	Optional: Set to block to reject, in enforce mode, queries the Loki guardrail cannot size up. Default `allow`.

In a container, a setting goes in with -e NAME=value after docker run.

What it does

Grafana Labs publishes this open-source server so an assistant can work inside a Grafana instance. It finds and edits dashboards, and it queries Prometheus, Loki, Tempo or Pyroscope through the datasources you already have. Alert rules and silences are in there too. So are Grafana Incident and OnCall, which matters if your on-call rota and incident timeline already live in Grafana and you want the assistant to update them while it digs through the data.

You run it yourself with uvx or the Docker image and point it at Grafana with GRAFANA_URL and a service account token. In the code GRAFANA_URL falls back to http://localhost:3000 when unset. What the tools can touch comes down to that token's RBAC permissions, and --disable-write turns writes off.

A good share of the tools stay off until you ask for them. The default --enabled-tools list leaves out seven datasource groups, SQL and CloudWatch among them. Admin tools and run_panel_query are opt-in as well. Since 2.2.0 the SQL tools also take BigQuery datasources, and CloudWatch has three Logs Insights tools next to the metrics ones. Listing log groups stops at 50, so on a big AWS account you pass a name prefix. A Logs Insights query gets 30 seconds to finish.

PromQL and LogQL queries can be held back by guardrails, which are off unless you set a mode. Put the Prometheus one in enforce mode and it turns away a query spanning more than 192 hours or asking for more than 11000 points per series, unless you change those caps. The Loki one has a 24 hour range cap by default. Shadow mode only logs what would have been blocked.

If your stack is on Grafana Cloud you can skip the install. Grafana runs a hosted copy at mcp.grafana.com that you sign in to with OAuth in the browser.

What it needs

To run it yourself	PythonDocker
Runs on	macOS, Windows, Linux
Grafana	Your own instanceSelf-managed Grafana 9.0 or later, or a Grafana Cloud stack, with a service account token
Also	uv on PATH for the uvx command

Create a service account in Grafana and give it the roles the tools need. The README lists them per tool, and the Editor role is the simpler but wider choice. Then make a token for it. The hosted endpoint needs a Grafana Cloud stack where the Grafana Assistant terms have been accepted. Your user also needs the Cloud MCP access permission, which Editors and above have by default.

Limits

Limit SQL, InfluxDB, CloudWatch, Cloud Logging, Elasticsearch, Quickwit, Graphite, admin and run_panel_query tools are off until named in --enabled-tools
Limit Under --disable-write, query_sql and query_influxdb are dropped too unless --enable-query is set, since they pass queries through unfiltered
Limit The Prometheus and Loki query guardrails are off by default and only reject queries once their mode is set to enforce
Size cap list_cloudwatch_log_groups returns at most 50 log groups, and query_cloudwatch_logs waits up to 30 seconds for a result
Size cap update_dashboard takes the whole dashboard JSON, which can fill a lot of the model's context on large dashboards
Limit Panel images need the Grafana Image Renderer installed on the Grafana side
Limit Cloud Logging tools need the Google Cloud Logging plugin 1.8.0 or later on Grafana 11.2 or later
Not covered The hosted endpoint works with Grafana Cloud stacks only and speaks streamable HTTP, with no SSE
Limit GRAFANA_API_KEY still works but is deprecated in favour of GRAFANA_SERVICE_ACCOUNT_TOKEN

Who it suits

Teams already on Grafana who want an assistant to read dashboards and traces during an incident. Run it locally for a self-managed instance, where the token and --disable-write keep its reach small. On Grafana Cloud the hosted endpoint saves the install and acts as the signed-in user. If you only need raw Prometheus or Loki access without Grafana in between, a server for that backend may be simpler.

Tools 60 of 123

47 only read, 11 change things, 2 can delete or overwrite.

search_dashboards: Find dashboards by title, folder, tag or starred state
get_dashboard_by_uid: Fetch a dashboard's JSON by UID, or one of its saved versions
list_dashboard_versions: List a dashboard's saved versions with author, time and message
get_dashboard_summary: Return a short outline of a dashboard instead of its full JSON
get_dashboard_property: Pull parts of a dashboard out with a JSONPath expression
get_dashboard_panel_queries: Show each panel's title, queries and datasource
update_dashboard changes: Create a dashboard or overwrite an existing one with new JSON
run_panel_query: Run the queries behind one or more dashboard panels (off by default)
create_folder changes: Create a dashboard folder
list_datasources: List the configured datasources
get_datasource: Fetch one datasource by UID or name
check_datasources_health: Run Grafana's health check on one or more datasources
create_datasource changes: Add a new datasource
update_datasource changes: Change an existing datasource's settings
query_prometheus: Run a PromQL instant or range query
list_prometheus_metric_names: List metric names in a Prometheus datasource
list_prometheus_label_names: List label names that match a selector
list_prometheus_label_values: List the values one label takes
query_prometheus_histogram: Work out histogram percentiles from Prometheus buckets
query_loki_logs: Run a LogQL log or metric query
list_loki_label_names: List the label names in a Loki datasource
list_loki_label_values: List the values of one Loki label
query_loki_stats: Report stream, chunk and byte counts for a log selector
query_loki_patterns: Return the log patterns Loki has detected
analyze_loki_labels: Review a Loki label setup and point out costly labels
query_influxdb changes: Run an InfluxQL or Flux query (off by default, can change data)
list_sql_databases: List databases, schemas, catalogs or BigQuery datasets in a SQL datasource (off by default)
list_sql_tables: List tables in a SQL datasource (off by default)
describe_sql_table: Show a table's columns and types (off by default)
query_sql changes: Run raw SQL through a Grafana SQL datasource, BigQuery included (off by default, can change data)
list_cloudwatch_namespaces: List AWS CloudWatch namespaces (off by default)
list_cloudwatch_metrics: List the metrics in a CloudWatch namespace (off by default)
list_cloudwatch_dimensions: List the dimensions of a CloudWatch metric (off by default)
query_cloudwatch: Run a CloudWatch metrics query over a time range (off by default)
list_cloudwatch_log_groups: List up to 50 CloudWatch log groups in a region, filtered by name prefix (off by default)
list_cloudwatch_log_group_fields: List the fields found in a CloudWatch log group (off by default)
query_cloudwatch_logs: Run a Logs Insights query over one or more log groups (off by default)
query_cloud_logging: Search Google Cloud Logging entries (off by default)
query_elasticsearch: Search Elasticsearch or OpenSearch with Lucene or Query DSL (off by default)
query_graphite: Fetch Graphite series for a target expression (off by default)
query_pyroscope: Fetch profiles or profile metrics from Pyroscope
search_tempo_traces: Find traces with a TraceQL query
get_tempo_trace: Fetch one trace by its ID
query_tempo_metrics: Compute TraceQL metrics over traces
list_incidents: List Grafana Incident incidents
get_incident: Fetch one incident with its custom fields
create_incident changes: Open a new incident in Grafana Incident
update_incident changes: Change an incident's status, severity, title or custom fields
alerting_rules_read: List alert rules, read one, or see its version history
alerting_rules_write deletes: Create, edit or delete alert rules
alerting_silences_read: List silences or read one
alerting_silences_write changes: Create, edit or expire silences
list_oncall_schedules: List Grafana OnCall schedules
get_current_oncall_users: Show who is on call now for a schedule
list_alert_groups: List OnCall alert groups with filters
update_alert_group changes: Acknowledge or resolve an OnCall alert group, or undo either
install_plugin changes: Install a Grafana plugin (write mode only)
render_grafana_app: Run an enabled query tool and show its result as an interactive Grafana view in hosts that support it
user_info: Show who the token belongs to, what it may do and which orgs it reaches
grafana_api_request deletes: Call any Grafana HTTP API path, GET only under --disable-write

Popularity

103 popularity score, on our list (the bar is 36)

★ 3.6k ↓ 740k/mo

GitHub stars	3,646	42.7 points
Stars in 30 days	+115	10 points
PyPI downloads, last month	740,122	42.1 points
Docker Hub pulls	5,153,100
Official server	Yes	8 points

Measured 2026-11-07. How the score works.

Where it is published

Code	grafana/mcp-grafana ★ 3.6k
PyPI	mcp-grafana 2.2.0, 2026-11-02, the one the command runs
Docker Hub	grafana/mcp-grafana
Hosted	`https://mcp.grafana.com/mcp` Streamable HTTP · OAuth sign-in

What it can reach

Runs	On your machine, or on the maker's servers
Reaches	The internet, Your account at a service, Cloud resources
Changes things	Yes: it can write, send, create or delete
Sign-in	OAuth sign-in for the hosted endpoint; an API key or token to run it yourself
Licence	Apache-2.0 (GitHub)
Usage data	Sends it unless you turn it off (`GRAFANA_USAGE_STATS=disabled`) Which tools ran, failure counts and the flag names in use, sent every 4 hours and at shutdown to stats.grafana.org. No arguments, queries or credentials
Last code change	2026-11-06

Runs on your machine or in your own container and calls the Grafana HTTP API with a service account token or basic auth, so it reaches what that account's RBAC allows. With writes on it can change dashboards, folders, datasources, alert rules, silences and incidents, install plugins, and send any method through grafana_api_request. Under --disable-write that tool allows GET plus read-only query POSTs, and query_sql and query_influxdb are removed unless --enable-query is set. Over SSE or streamable HTTP, set MCP_GRAFANA_SERVER_TOKEN so clients must present a bearer token. The hosted endpoint runs on Grafana's servers, signs in with OAuth and acts as the signed-in user.

Over time

GitHub stars 3.6k +115 (+3 %)

3.6k3.5k

8 Oct 20267 Nov 2026

PyPI downloads a month 740k +514k (+227 %)

772k212k

12 May 20267 Nov 2026

Before the site measured it, read from pypistats' daily download counts.

Popularity score 103 +24 (+30 %)

10379

8 Oct 20267 Nov 2026

What is rising across the directory

Sources

Page checked 2026-11-07.

Badge for your README

Shows this server's rank in its category, and stars for a server that runs on your machine. It updates every hour.

MarkdownREADME.md

[![Grafana MCP server on EveryMCPServer](https://everymcpserver.com/badge/mcp-grafana-2.svg)](https://everymcpserver.com/servers/mcp-grafana-2)

Connect your client

Add in one click

Or add it by command or config

Add in one click

Or add it by command or config

Add in one click

Or add it by command or config

Settings 12

What it does

What it needs

Limits

Who it suits

Tools 60 of 123

Similar servers

Databuddy MCP server

PostHog MCP Server

Better Stack MCP Server

Coralogix MCP Server

Sentry MCP

InfluxDB 3 MCP Server