Grafana Cloud Mcp Tools

作者 grafana12be5634a492無授權條款18 個星標收錄於 2026年10月8日更新於 2026年10月8日儲存庫今天更新

Connect to and use the Grafana Cloud MCP server effectively, including natural-language Grafana Assistant questions and observability investigations. Covers setup, OAuth authorization, tool categories, read/write access scopes, and best practices for context window management. Use when the user wants to set up Grafana Cloud MCP, ask Grafana Assistant a question, investigate metrics, logs, traces, profiles, dashboards, alerts, or incidents, needs guidance on which tool to use, or wants to understand access scopes.

AI 產生的概覽

說明如何設定與使用託管的 Grafana Cloud MCP 伺服器,以進行可觀測性查詢與調查。

功能
說明如何透過 Streamable HTTP 與 OAuth 2.1 授權,將代理連線至託管的 Grafana Cloud MCP 伺服器,涵蓋先決條件、設定步驟以及讀寫存取範圍。內容列出伺服器的工具類別,包括儀表板、資料來源、Prometheus、Loki、Tempo、Pyroscope、ClickHouse、CloudWatch、Elasticsearch、告警、註解、事件、OnCall 與 Sift。同時提供自然語言 Grafana Assistant 提問、上下文視窗管理、查詢,以及安全使用寫入操作的最佳實務。
適用情境
適用於設定 Grafana Cloud MCP 存取、向 Grafana Assistant 提問,或調查指標、日誌、追蹤、效能剖析、儀表板、告警或事件時。也有助於選擇應呼叫哪個 Grafana 工具,或釐清存取範圍與權限。
執行需求
需要具備 Assistant Cloud MCP User 角色或同等權限的 Grafana Cloud 帳號,並需由管理員接受 Grafana Assistant 條款。需要連線至託管 MCP 端點的網路存取,以及瀏覽器式 OAuth 授權;寫入工具需要 grafana:write 範圍,以及 Assistant Admin 角色或同等權限。未包含本機安裝、環境變數或指令碼。

Grafana Cloud MCP Server

Hosted MCP server providing 60+ Grafana Cloud tools via Streamable HTTP transport with OAuth 2.1 authorization.

Prerequisites

  • Grafana Cloud account (self-hosted users should use the local OSS MCP server instead)
  • Assistant Cloud MCP User role or grafana-assistant-app.cloud-mcp:access permission (Editor role or higher has this by default)
  • An administrator must have accepted the Grafana Assistant terms and conditions

Configuration

The Cloud MCP server connects via https://mcp.grafana.com/mcp using Streamable HTTP transport. No local installation or environment variables are required.

Setup steps

  1. The MCP server connects automatically when the plugin loads.
  2. When prompted, enter your Grafana Cloud URL and authorize the connection in your browser.
  3. Your OAuth token is valid for 1 hour and refreshes automatically for 30 days.

Read and write access

When authorizing, you choose which permissions to grant:

  • Read access: View dashboards, alerts, incidents, and query data sources. Always available.
  • Write access: Create and modify dashboards, alerts, and incidents. You can uncheck this to grant read-only access.

Organization Admins can grant write access by default. If write access is disabled on the consent page, the user needs the Assistant Admin role.

Tool categories

Search and navigation

  • search_dashboards — search for dashboards by query string
  • search_folders — search for folders by query string
  • generate_deeplink — generate deeplink URLs for dashboards, panels, and Explore queries

Dashboards and folders

  • get_dashboard_by_uid — retrieve the complete dashboard JSON by UID
  • get_dashboard_summary — compact summary without full JSON (preferred)
  • get_dashboard_property — extract specific parts via JSONPath
  • get_dashboard_panel_queries — retrieve panel queries with template variable substitution
  • update_dashboard — create or update a dashboard (Write)
  • create_folder — create a Grafana folder (Write)

Datasources

  • list_datasources — list all configured data sources with optional type filtering
  • get_datasource — get detailed information by UID or name

Prometheus

  • list_prometheus_metric_names — discover available metrics with regex filtering
  • list_prometheus_metric_metadata — list metadata about currently scraped metrics
  • list_prometheus_label_names — list label names with optional series selector
  • list_prometheus_label_values — get values for a specific label
  • query_prometheus — execute PromQL instant or range queries
  • query_prometheus_histogram — query histogram percentiles

Loki

  • list_loki_label_names — list available label names in logs
  • list_loki_label_values — get unique values for a specific label
  • query_loki_logs — execute LogQL queries for log entries or metric values
  • query_loki_stats — get statistics about log streams
  • query_loki_patterns — detect and analyze common log patterns

Tempo

If you have Tempo data sources, the Cloud MCP server proxies tools from the Tempo data source, including TraceQL queries and attribute discovery.

Pyroscope

  • list_pyroscope_label_names — list available label names in profiles
  • list_pyroscope_label_values — list values for a specific label
  • list_pyroscope_profile_types — list available profile types
  • query_pyroscope — query profiles or metrics from Pyroscope

ClickHouse

  • list_clickhouse_tables — list available tables with metadata
  • describe_clickhouse_table — get column schema for a table
  • query_clickhouse — execute SQL queries against ClickHouse datasources

CloudWatch

  • list_cloudwatch_namespaces — list available AWS namespaces
  • list_cloudwatch_metrics — list metrics for a namespace
  • list_cloudwatch_dimensions — list dimension keys for a metric
  • query_cloudwatch — query AWS CloudWatch metrics

Elasticsearch

  • query_elasticsearch — execute Lucene or Query DSL searches

Alerting

  • alerting_manage_rules — list, filter, create, and update alert rules (Read / Write)
  • alerting_manage_routing — view routing configuration, notification policies, contact points (Read)

Annotations

  • get_annotations — fetch annotations filtered by dashboard UID, time range, or tags
  • get_annotation_tags — get annotation tags with optional filtering
  • create_annotation — create a new annotation (Write)
  • update_annotation — update an existing annotation (Write)

Incidents

  • list_incidents — list incidents with optional status filtering
  • get_incident — get full incident details by ID
  • create_incident — create a new incident (Write)
  • add_activity_to_incident — add a note to an incident's timeline (Write)

OnCall

  • list_oncall_schedules — list OnCall schedules with optional team filtering
  • get_oncall_shift — get detailed shift information
  • get_current_oncall_users — get users currently on-call for a schedule
  • list_oncall_teams — list OnCall teams
  • list_oncall_users — list OnCall users with optional filtering
  • list_alert_groups — list alert groups with filtering
  • get_alert_group — get a specific alert group by ID

Sift

  • list_sift_investigations — list Sift investigations
  • get_sift_investigation — retrieve a Sift investigation by UUID
  • get_sift_analysis — retrieve a specific analysis from an investigation
  • find_error_pattern_logs — search Loki logs for elevated error patterns (Write)
  • find_slow_requests — search Tempo datasources for slow requests (Write)

Other tools

  • ask_assistant — send a prompt to Grafana Assistant and get the full reply (Write)
  • get_assertions — get assertion summary for an entity
  • get_panel_image — render a dashboard panel as a PNG image
  • describe_infrastructure — retrieve pre-built summaries of service groups
  • get_query_examples — get example queries for datasource types

Best practices

Grafana Assistant questions

  • For open-ended, natural-language requests ("Why is checkout latency spiking?", "Summarize the current incident", "Explain this alert"), use ask_assistant to let Grafana Assistant plan the investigation and return a grounded reply. ask_assistant is write-scoped and requires grafana:write.
  • If write access was not granted during OAuth authorization, ask_assistant is unavailable. Fall back to the appropriate read-only tools (for example describe_infrastructure, get_assertions, query_prometheus, query_loki_logs, list_incidents) and tell the user that the natural-language Assistant tool needs write scope. Re-authorizing only grants write if the user has the Assistant Admin role or the grafana-assistant-app.cloud-mcp.scope:write permission; otherwise a Grafana organization administrator must grant it first.
  • For specific, targeted lookups (a known query, dashboard, alert, or incident), call the corresponding tool directly rather than routing through ask_assistant.

Context window management

  • Use get_dashboard_summary instead of get_dashboard_by_uid to avoid consuming context with full dashboard JSON.
  • Use get_dashboard_property with JSONPath to extract only the specific parts you need.
  • Use search_dashboards to discover dashboards before retrieving by UID.
  • When presenting Grafana data, use generate_deeplink to provide clickable URLs rather than describing navigation steps.

Querying

  • When querying Prometheus, always specify a reasonable time range to avoid overwhelming results.
  • When querying Loki, prefer targeted LogQL selectors with label matchers over broad queries.
  • Use datasource discovery tools (list_datasources, list_prometheus_metric_names) before writing queries.

Safety

  • Avoid write operations (update_dashboard, create_incident, alerting_manage_rules) unless explicitly asked by the user.
  • If write access was not granted during OAuth authorization, write tools will not be available.

Access and permissions

  • Access is user-scoped: the agent has only the permissions your Grafana RBAC grants you.
  • Using the Cloud MCP server requires the Assistant Cloud MCP User role or the grafana-assistant-app.cloud-mcp:access permission. Editor and higher have this by default; other roles can be granted it explicitly.
  • Write tools require the grafana:write OAuth scope, granted during OAuth consent. Granting that scope is itself gated by Grafana RBAC: the user needs the Assistant Admin role or the grafana-assistant-app.cloud-mcp.scope:write permission to authorize a write-scoped connection. Read-only access does not require it.

來源與署名

來源:grafana/ai-marketplace位於plugins/grafana-cloud-mcp/skills/grafana-cloud-mcp-tools提交12be563

授權條款: 無授權條款

內容歸原作者所有。SourceWeft 從公開儲存庫中收錄這些內容。

檢舉或申請下架